Change -V option to -v everywhere

This might occur if things change while we're calculating the sizes.

.coveragerc

@@ -7,4 +7,4 @@
 exclude_lines =
 	pragma: no cover
 	if 0:
-omit = nilmdb/utils/datetime_tz*,nilmdb/scripts,nilmdb/_version.py
+omit = nilmdb/utils/datetime_tz*,nilmdb/scripts,nilmdb/_version.py,nilmdb/fsck

README.txt

@@ -8,7 +8,8 @@ Prerequisites:
 
   # Base NilmDB dependencies
   sudo apt-get install python-cherrypy3 python-decorator python-simplejson
-  sudo apt-get install python-requests python-dateutil python-tz python-psutil
+  sudo apt-get install python-requests python-dateutil python-tz
+  sudo apt-get install python-progressbar python-psutil
 
   # Other dependencies (required by some modules)
   sudo apt-get install python-numpy
@@ -26,6 +27,7 @@ Install:
 Usage:
 
   nilmdb-server --help
+  nilmdb-fsck --help
   nilmtool --help
 
 See docs/wsgi.md for info on setting up a WSGI application in Apache.

docs/design.md

@@ -421,3 +421,20 @@ and has all of the same functions.  It adds three new functions:
 It is significantly faster!  It is about 20 times faster to decimate a
 stream with `nilm-decimate` when the filter code is using the new
 binary/numpy interface.
+
+
+WSGI interface & chunked requests
+---------------------------------
+
+mod_wsgi requires "WSGIChunkedRequest On" to handle
+"Transfer-encoding: Chunked" requests.  However, `/stream/insert`
+doesn't handle this correctly right now, because:
+
+- The `cherrpy.request.body.read()` call needs to be fixed for chunked requests
+
+- We don't want to just buffer endlessly in the server, and it will
+  require some thought on how to handle data in chunks (what to do about
+  interval endpoints).
+
+It is probably better to just keep the endpoint management on the client
+side, so leave "WSGIChunkedRequest off" for now.

docs/wsgi.md

@@ -19,12 +19,12 @@ Then, set up Apache with a configuration like:
 
     <VirtualHost>
         WSGIScriptAlias /nilmdb /home/nilm/nilmdb.wsgi
-        WSGIApplicationGroup nilmdb-appgroup
-        WSGIProcessGroup nilmdb-procgroup
         WSGIDaemonProcess nilmdb-procgroup threads=32 user=nilm group=nilm
-
-        # Access control example:
         <Location /nilmdb>
+            WSGIProcessGroup nilmdb-procgroup
+            WSGIApplicationGroup nilmdb-appgroup
+
+            # Access control example:
             Order deny,allow
             Deny from all
             Allow from 1.2.3.4

extras/fix-oversize-files.py

@@ -0,0 +1,50 @@
+#!/usr/bin/python
+
+import os
+import sys
+import cPickle as pickle
+import argparse
+import fcntl
+import re
+from nilmdb.client.numpyclient import layout_to_dtype
+
+parser = argparse.ArgumentParser(
+    description = """
+Fix database corruption where binary writes caused too much data to be
+written to the file.  Truncates files to the correct length.  This was
+fixed by b98ff1331a515ad47fd3203615e835b529b039f9.
+""")
+parser.add_argument("path", action="store", help='Database root path')
+parser.add_argument("-y", "--yes", action="store_true", help='Fix them')
+args = parser.parse_args()
+
+lock = os.path.join(args.path, "data.lock")
+with open(lock, "w") as f:
+    fcntl.flock(f.fileno(), fcntl.LOCK_EX | fcntl.LOCK_NB)
+
+    fix = {}
+
+    for (path, dirs, files) in os.walk(args.path):
+        if "_format" in files:
+            with open(os.path.join(path, "_format")) as format:
+                fmt = pickle.load(format)
+                rowsize = layout_to_dtype(fmt["layout"]).itemsize
+                maxsize = rowsize * fmt["rows_per_file"]
+                fix[path] = maxsize
+                if maxsize < 128000000: # sanity check
+                    raise Exception("bad maxsize " + str(maxsize))
+
+    for fixpath in fix:
+        for (path, dirs, files) in os.walk(fixpath):
+            for fn in files:
+                if not re.match("^[0-9a-f]{4,}$", fn):
+                    continue
+                fn = os.path.join(path, fn)
+                size = os.path.getsize(fn)
+                maxsize = fix[fixpath]
+                if size > maxsize:
+                    diff = size - maxsize
+                    print diff, "too big:", fn
+                    if args.yes:
+                        with open(fn, "a+") as dbfile:
+                            dbfile.truncate(maxsize)

nilmdb/client/client.py

@@ -6,7 +6,6 @@ import nilmdb.utils
 import nilmdb.client.httpclient
 from nilmdb.client.errors import ClientError
 
-import re
 import time
 import simplejson as json
 import contextlib
@@ -59,6 +58,11 @@ class Client(object):
         return self.http.get("dbinfo")
 
     def stream_list(self, path = None, layout = None, extended = False):
+        """Return a sorted list of [path, layout] lists.  If 'path' or
+        'layout' are specified, only return streams that match those
+        exact values.  If 'extended' is True, the returned lists have
+        extended info, e.g.: [path, layout, extent_min, extent_max,
+        total_rows, total_seconds."""
         params = {}
         if path is not None:
             params["path"] = path
@@ -66,14 +70,11 @@ class Client(object):
             params["layout"] = layout
         if extended:
             params["extended"] = 1
-        def sort_streams_nicely(x):
-            """Human-friendly sort (/stream/2 before /stream/10)"""
-            num = lambda t: int(t) if t.isdigit() else t
-            key = lambda k: [ num(c) for c in re.split('([0-9]+)', k[0]) ]
-            return sorted(x, key = key)
-        return sort_streams_nicely(self.http.get("stream/list", params))
+        streams = self.http.get("stream/list", params)
+        return nilmdb.utils.sort.sort_human(streams, key = lambda s: s[0])
 
     def stream_get_metadata(self, path, keys = None):
+        """Get stream metadata"""
         params = { "path": path }
         if keys is not None:
             params["key"] = keys
@@ -122,7 +123,10 @@ class Client(object):
             params["start"] = timestamp_to_string(start)
         if end is not None:
             params["end"] = timestamp_to_string(end)
-        return self.http.post("stream/remove", params)
+        total = 0
+        for count in self.http.post_gen("stream/remove", params):
+            total += int(count)
+        return total
 
     @contextlib.contextmanager
     def stream_insert_context(self, path, start = None, end = None):
@@ -146,6 +150,7 @@ class Client(object):
         ctx = StreamInserter(self, path, start, end)
         yield ctx
         ctx.finalize()
+        ctx.destroy()
 
     def stream_insert(self, path, data, start = None, end = None):
         """Insert rows of data into a stream.  data should be a string
@@ -295,6 +300,15 @@ class StreamInserter(object):
         self._block_data = []
         self._block_len = 0
 
+        self.destroyed = False
+
+    def destroy(self):
+        """Ensure this object can't be used again without raising
+        an error"""
+        def error(*args, **kwargs):
+            raise Exception("don't reuse this context object")
+        self._send_block = self.insert = self.finalize = self.send = error
+
     def insert(self, data):
         """Insert a chunk of ASCII formatted data in string form.  The
         overall data must consist of lines terminated by '\\n'."""
@@ -441,7 +455,7 @@ class StreamInserter(object):
             self._interval_start = end_ts
 
         # Double check endpoints
-        if start_ts is None or end_ts is None:
+        if (start_ts is None or end_ts is None) or (start_ts == end_ts):
             # If the block has no non-comment lines, it's OK
             try:
                 self._get_first_noncomment(block)

nilmdb/client/httpclient.py

@@ -9,7 +9,7 @@ import requests
 
 class HTTPClient(object):
     """Class to manage and perform HTTP requests from the client"""
-    def __init__(self, baseurl = "", post_json = False):
+    def __init__(self, baseurl = "", post_json = False, verify_ssl = True):
         """If baseurl is supplied, all other functions that take
         a URL can be given a relative URL instead."""
         # Verify / clean up URL
@@ -19,6 +19,7 @@ class HTTPClient(object):
         self.baseurl = reparsed.rstrip('/') + '/'
 
         # Build Requests session object, enable SSL verification
+        self.verify_ssl = verify_ssl
         self.session = requests.Session()
         self.session.verify = True
 
@@ -67,7 +68,8 @@ class HTTPClient(object):
                                             params = query_data,
                                             data = body_data,
                                             stream = stream,
-                                            headers = headers)
+                                            headers = headers,
+                                            verify = self.verify_ssl)
         except requests.RequestException as e:
             raise ServerError(status = "502 Error", url = url,
                               message = str(e.message))
@@ -123,19 +125,50 @@ class HTTPClient(object):
         """
         (response, isjson) = self._do_req(method, url, query, body,
                                           stream = True, headers = headers)
+
+        # Like the iter_lines function in Requests, but only splits on
+        # the specified line ending.
+        def lines(source, ending):
+            pending = None
+            for chunk in source:
+                if pending is not None:
+                    chunk = pending + chunk
+                tmp = chunk.split(ending)
+                lines = tmp[:-1]
+                if chunk.endswith(ending):
+                    pending = None
+                else:
+                    pending = tmp[-1]
+                for line in lines:
+                    yield line
+            if pending is not None: # pragma: no cover (missing newline)
+                yield pending
+
+        # Yield the chunks or lines as requested
         if binary:
             for chunk in response.iter_content(chunk_size = 65536):
                 yield chunk
         elif isjson:
-            for line in response.iter_lines():
+            for line in lines(response.iter_content(chunk_size = 1),
+                              ending = '\r\n'):
                 yield json.loads(line)
         else:
-            for line in response.iter_lines():
+            for line in lines(response.iter_content(chunk_size = 65536),
+                              ending = '\n'):
                 yield line
 
     def get_gen(self, url, params = None, binary = False):
         """Simple GET (parameters in URL) returning a generator"""
         return self._req_gen("GET", url, params, binary = binary)
 
+    def post_gen(self, url, params = None):
+        """Simple POST (parameters in body) returning a generator"""
+        if self.post_json:
+            return self._req_gen("POST", url, None,
+                                 json.dumps(params),
+                                 { 'Content-type': 'application/json' })
+        else:
+            return self._req_gen("POST", url, None, params)
+
     # Not much use for a POST or PUT generator, since they don't
     # return much data.

nilmdb/client/numpyclient.py

@@ -98,6 +98,7 @@ class NumpyClient(nilmdb.client.client.Client):
         ctx = StreamInserterNumpy(self, path, start, end, dtype)
         yield ctx
         ctx.finalize()
+        ctx.destroy()
 
     def stream_insert_numpy(self, path, data, start = None, end = None,
                             layout = None):
@@ -133,16 +134,8 @@ class StreamInserterNumpy(nilmdb.client.client.StreamInserter):
         contiguous interval and may be None.  'dtype' is the Numpy
         dtype for this stream.
         """
-        self.last_response = None
-
+        super(StreamInserterNumpy, self).__init__(client, path, start, end)
         self._dtype = dtype
-        self._client = client
-        self._path = path
-
-        # Start and end for the overall contiguous interval we're
-        # filling
-        self._interval_start = start
-        self._interval_end = end
 
         # Max rows to send at once
         self._max_rows = self._max_data // self._dtype.itemsize
@@ -162,9 +155,12 @@ class StreamInserterNumpy(nilmdb.client.client.StreamInserter):
         elif array.ndim == 2:
             # Convert to structured array
             sarray = numpy.zeros(array.shape[0], dtype=self._dtype)
-            sarray['timestamp'] = array[:,0]
-            # Need the squeeze in case sarray['data'] is 1 dimensional
-            sarray['data'] = numpy.squeeze(array[:,1:])
+            try:
+                sarray['timestamp'] = array[:,0]
+                # Need the squeeze in case sarray['data'] is 1 dimensional
+                sarray['data'] = numpy.squeeze(array[:,1:])
+            except (IndexError, ValueError):
+                raise ValueError("wrong number of fields for this data type")
             array = sarray
         else:
             raise ValueError("wrong number of dimensions in array")
@@ -247,9 +243,12 @@ class StreamInserterNumpy(nilmdb.client.client.StreamInserter):
             # Next block continues where this one ended
             self._interval_start = end_ts
 
-        # If we have no endpoints, it's because we had no data to send.
-        if start_ts is None or end_ts is None:
-            return
+        # If we have no endpoints, or equal endpoints, it's OK as long
+        # as there's no data to send
+        if (start_ts is None or end_ts is None) or (start_ts == end_ts):
+            if len(array) == 0:
+                return
+            raise ClientError("have data to send, but invalid start/end times")
 
         # Send it
         data = array.tostring()

nilmdb/cmdline/cmdline.py

@@ -19,9 +19,8 @@ except ImportError: # pragma: no cover
 
 # Valid subcommands.  Defined in separate files just to break
 # things up -- they're still called with Cmdline as self.
-subcommands = [ "help", "info", "create", "list", "metadata",
-                "insert", "extract", "remove", "destroy",
-                "intervals", "rename" ]
+subcommands = [ "help", "info", "create", "rename", "list", "intervals",
+                "metadata", "insert", "extract", "remove", "destroy" ]
 
 # Import the subcommand modules
 subcmd_mods = {}
@@ -29,6 +28,14 @@ for cmd in subcommands:
     subcmd_mods[cmd] = __import__("nilmdb.cmdline." + cmd, fromlist = [ cmd ])
 
 class JimArgumentParser(argparse.ArgumentParser):
+    def parse_args(self, args=None, namespace=None):
+        # Look for --version anywhere and change it to just "nilmtool
+        # --version".  This makes "nilmtool cmd --version" work, which
+        # is needed by help2man.
+        if "--version" in (args or sys.argv[1:]):
+            args = [ "--version" ]
+        return argparse.ArgumentParser.parse_args(self, args, namespace)
+
     def error(self, message):
         self.print_usage(sys.stderr)
         self.exit(2, sprintf("error: %s\n", message))
@@ -72,15 +79,27 @@ class Complete(object): # pragma: no cover
         path = parsed_args.path
         if not path:
             return []
-        return ( self.escape(k + '=' + v)
-                 for (k,v) in client.stream_get_metadata(path).iteritems()
-                 if k.startswith(prefix) )
-
+        results = []
+        # prefix comes in as UTF-8, but results need to be Unicode,
+        # weird.  Still doesn't work in all cases, but that's bugs in
+        # argcomplete.
+        prefix = nilmdb.utils.unicode.decode(prefix)
+        for (k,v) in client.stream_get_metadata(path).iteritems():
+            kv = self.escape(k + '=' + v)
+            if kv.startswith(prefix):
+                results.append(kv)
+        return results
 
 class Cmdline(object):
 
     def __init__(self, argv = None):
         self.argv = argv or sys.argv[1:]
+        try:
+            # Assume command line arguments are encoded with stdin's encoding,
+            # and reverse it.  Won't be needed in Python 3, but for now..
+            self.argv = [ x.decode(sys.stdin.encoding) for x in self.argv ]
+        except Exception: # pragma: no cover
+            pass
         self.client = None
         self.def_url = os.environ.get("NILMDB_URL", "http://localhost/nilmdb/")
         self.subcmd = {}
@@ -102,7 +121,7 @@ class Cmdline(object):
         group = self.parser.add_argument_group("General options")
         group.add_argument("-h", "--help", action='help',
                            help='show this help message and exit')
-        group.add_argument("-V", "--version", action="version",
+        group.add_argument("-v", "--version", action="version",
                            version = nilmdb.__version__)
 
         group = self.parser.add_argument_group("Server")

nilmdb/cmdline/destroy.py

@@ -1,5 +1,6 @@
 from nilmdb.utils.printf import *
 import nilmdb.client
+import fnmatch
 
 from argparse import ArgumentDefaultsHelpFormatter as def_form
 
@@ -10,25 +11,39 @@ def setup(self, sub):
                          Destroy the stream at the specified path.
                          The stream must be empty.  All metadata
                          related to the stream is permanently deleted.
+
+                         Wildcards and multiple paths are supported.
                          """)
     cmd.set_defaults(handler = cmd_destroy)
     group = cmd.add_argument_group("Options")
     group.add_argument("-R", "--remove", action="store_true",
                        help="Remove all data before destroying stream")
+    group.add_argument("-q", "--quiet", action="store_true",
+                       help="Don't display names when destroying "
+                       "multiple paths")
     group = cmd.add_argument_group("Required arguments")
-    group.add_argument("path",
-                       help="Path of the stream to delete, e.g. /foo/bar",
+    group.add_argument("path", nargs='+',
+                       help="Path of the stream to delete, e.g. /foo/bar/*",
                        ).completer = self.complete.path
     return cmd
 
 def cmd_destroy(self):
     """Destroy stream"""
-    if self.args.remove:
+    streams = [ s[0] for s in self.client.stream_list() ]
+    paths = []
+    for path in self.args.path:
+        new = fnmatch.filter(streams, path)
+        if not new:
+            self.die("error: no stream matched path: %s", path)
+        paths.extend(new)
+
+    for path in paths:
+        if not self.args.quiet and len(paths) > 1:
+            printf("Destroying %s\n", path)
+
         try:
-            count = self.client.stream_remove(self.args.path)
+            if self.args.remove:
+                count = self.client.stream_remove(path)
+            self.client.stream_destroy(path)
         except nilmdb.client.ClientError as e:
-            self.die("error removing data: %s", str(e))
-    try:
-        self.client.stream_destroy(self.args.path)
-    except nilmdb.client.ClientError as e:
-        self.die("error destroying stream: %s", str(e))
+            self.die("error destroying stream: %s", str(e))

nilmdb/cmdline/extract.py

@@ -1,6 +1,7 @@
 from __future__ import print_function
 from nilmdb.utils.printf import *
 import nilmdb.client
+import sys
 
 def setup(self, sub):
     cmd = sub.add_parser("extract", help="Extract data",
@@ -24,6 +25,8 @@ def setup(self, sub):
                        ).completer = self.complete.time
 
     group = cmd.add_argument_group("Output format")
+    group.add_argument("-B", "--binary", action="store_true",
+                       help="Raw binary output")
     group.add_argument("-b", "--bare", action="store_true",
                        help="Exclude timestamps from output lines")
     group.add_argument("-a", "--annotate", action="store_true",
@@ -42,6 +45,11 @@ def cmd_extract_verify(self):
         if self.args.start > self.args.end:
             self.parser.error("start is after end")
 
+    if self.args.binary:
+        if (self.args.bare or self.args.annotate or self.args.markup or
+            self.args.timestamp_raw or self.args.count):
+            self.parser.error("--binary cannot be combined with other options")
+
 def cmd_extract(self):
     streams = self.client.stream_list(self.args.path)
     if len(streams) != 1:
@@ -60,16 +68,23 @@ def cmd_extract(self):
         printf("# end: %s\n", time_string(self.args.end))
 
     printed = False
+    if self.args.binary:
+        printer = sys.stdout.write
+    else:
+        printer = print
+    bare = self.args.bare
+    count = self.args.count
     for dataline in self.client.stream_extract(self.args.path,
                                                self.args.start,
                                                self.args.end,
                                                self.args.count,
-                                               self.args.markup):
-        if self.args.bare and not self.args.count:
+                                               self.args.markup,
+                                               self.args.binary):
+        if bare and not count:
             # Strip timestamp (first element).  Doesn't make sense
             # if we are only returning a count.
             dataline = ' '.join(dataline.split(' ')[1:])
-        print(dataline)
+        printer(dataline)
         printed = True
     if not printed:
         if self.args.annotate:

nilmdb/cmdline/info.py

@@ -21,5 +21,8 @@ def cmd_info(self):
     printf("Server URL: %s\n", self.client.geturl())
     dbinfo = self.client.dbinfo()
     printf("Server database path: %s\n", dbinfo["path"])
-    printf("Server database size: %s\n", human_size(dbinfo["size"]))
-    printf("Server database free space: %s\n", human_size(dbinfo["free"]))
+    for (desc, field) in [("used by NilmDB", "size"),
+                          ("used by other", "other"),
+                          ("reserved", "reserved"),
+                          ("free", "free")]:
+        printf("Server disk space %s: %s\n", desc, human_size(dbinfo[field]))

nilmdb/cmdline/intervals.py

@@ -1,5 +1,6 @@
 from nilmdb.utils.printf import *
 import nilmdb.utils.time
+from nilmdb.utils.interval import Interval
 
 import fnmatch
 import argparse
@@ -42,6 +43,8 @@ def setup(self, sub):
     group = cmd.add_argument_group("Misc options")
     group.add_argument("-T", "--timestamp-raw", action="store_true",
                        help="Show raw timestamps when printing times")
+    group.add_argument("-o", "--optimize", action="store_true",
+                       help="Optimize (merge adjacent) intervals")
 
     return cmd
 
@@ -58,9 +61,16 @@ def cmd_intervals(self):
         time_string = nilmdb.utils.time.timestamp_to_human
 
     try:
-           for (start, end) in self.client.stream_intervals(
-               self.args.path, self.args.start, self.args.end, self.args.diff):
-               printf("[ %s -> %s ]\n", time_string(start), time_string(end))
+        intervals = ( Interval(start, end) for (start, end) in
+                      self.client.stream_intervals(self.args.path,
+                                                   self.args.start,
+                                                   self.args.end,
+                                                   self.args.diff) )
+        if self.args.optimize:
+            intervals = nilmdb.utils.interval.optimize(intervals)
+        for i in intervals:
+            printf("[ %s -> %s ]\n", time_string(i.start), time_string(i.end))
+
     except nilmdb.client.ClientError as e:
         self.die("error listing intervals: %s", str(e))
 

nilmdb/cmdline/list.py

@@ -10,22 +10,16 @@ def setup(self, sub):
                          formatter_class = def_form,
                          description="""
                          List streams available in the database,
-                         optionally filtering by layout or path.  Wildcards
-                         are accepted.
+                         optionally filtering by path.  Wildcards
+                         are accepted; non-matching paths or wildcards
+                         are ignored.
                          """)
     cmd.set_defaults(verify = cmd_list_verify,
                      handler = cmd_list)
 
     group = cmd.add_argument_group("Stream filtering")
-    group.add_argument("-p", "--path", metavar="PATH", default="*",
-                       help="Match only this path (-p can be omitted)",
+    group.add_argument("path", metavar="PATH", default=["*"], nargs='*',
                        ).completer = self.complete.path
-    group.add_argument("path_positional", default="*",
-                       nargs="?", help=argparse.SUPPRESS,
-                       ).completer = self.complete.path
-    group.add_argument("-l", "--layout", default="*",
-                       help="Match only this stream layout",
-                       ).completer = self.complete.layout
 
     group = cmd.add_argument_group("Interval info")
     group.add_argument("-E", "--ext", action="store_true",
@@ -49,20 +43,12 @@ def setup(self, sub):
     group = cmd.add_argument_group("Misc options")
     group.add_argument("-T", "--timestamp-raw", action="store_true",
                        help="Show raw timestamps when printing times")
+    group.add_argument("-l", "--layout", action="store_true",
+                       help="Show layout type next to path name")
 
     return cmd
 
 def cmd_list_verify(self):
-    # A hidden "path_positional" argument lets the user leave off the
-    # "-p" when specifying the path.  Handle it here.
-    got_opt = self.args.path != "*"
-    got_pos = self.args.path_positional != "*"
-    if got_pos:
-        if got_opt:
-            self.parser.error("too many paths specified")
-        else:
-            self.args.path = self.args.path_positional
-
     if self.args.start is not None and self.args.end is not None:
         if self.args.start >= self.args.end:
             self.parser.error("start must precede end")
@@ -80,29 +66,33 @@ def cmd_list(self):
     else:
         time_string = nilmdb.utils.time.timestamp_to_human
 
-    for stream in streams:
-        (path, layout, int_min, int_max, rows, time) = stream[:6]
-        if not (fnmatch.fnmatch(path, self.args.path) and
-                fnmatch.fnmatch(layout, self.args.layout)):
-            continue
+    for argpath in self.args.path:
+        for stream in streams:
+            (path, layout, int_min, int_max, rows, time) = stream[:6]
+            if not fnmatch.fnmatch(path, argpath):
+                continue
 
-        printf("%s %s\n", path, layout)
-
-        if self.args.ext:
-            if int_min is None or int_max is None:
-                printf("  interval extents: (no data)\n")
+            if self.args.layout:
+                printf("%s %s\n", path, layout)
             else:
-                printf("  interval extents: %s -> %s\n",
-                       time_string(int_min), time_string(int_max))
-            printf("        total data: %d rows, %.6f seconds\n",
-                   rows or 0,
-                   nilmdb.utils.time.timestamp_to_seconds(time or 0))
+                printf("%s\n", path)
 
-        if self.args.detail:
-            printed = False
-            for (start, end) in self.client.stream_intervals(
-                path, self.args.start, self.args.end):
-                printf("  [ %s -> %s ]\n", time_string(start), time_string(end))
-                printed = True
-            if not printed:
-                printf("  (no intervals)\n")
+            if self.args.ext:
+                if int_min is None or int_max is None:
+                    printf("  interval extents: (no data)\n")
+                else:
+                    printf("  interval extents: %s -> %s\n",
+                           time_string(int_min), time_string(int_max))
+                printf("        total data: %d rows, %.6f seconds\n",
+                       rows or 0,
+                       nilmdb.utils.time.timestamp_to_seconds(time or 0))
+
+            if self.args.detail:
+                printed = False
+                for (start, end) in self.client.stream_intervals(
+                    path, self.args.start, self.args.end):
+                    printf("  [ %s -> %s ]\n",
+                           time_string(start), time_string(end))
+                    printed = True
+                if not printed:
+                    printf("  (no intervals)\n")

nilmdb/cmdline/metadata.py

@@ -41,10 +41,10 @@ def cmd_metadata(self):
     if self.args.set is not None or self.args.update is not None:
         # Either set, or update
         if self.args.set is not None:
-            keyvals = self.args.set
+            keyvals = map(nilmdb.utils.unicode.decode, self.args.set)
             handler = self.client.stream_set_metadata
         else:
-            keyvals = self.args.update
+            keyvals = map(nilmdb.utils.unicode.decode, self.args.update)
             handler = self.client.stream_update_metadata
 
         # Extract key=value pairs
@@ -62,7 +62,9 @@ def cmd_metadata(self):
             self.die("error setting/updating metadata: %s", str(e))
     elif self.args.delete is not None:
         # Delete (by setting values to empty strings)
-        keys = self.args.delete or None
+        keys = None
+        if self.args.delete:
+            keys = map(nilmdb.utils.unicode.decode, self.args.delete)
         try:
             data = self.client.stream_get_metadata(self.args.path, keys)
             for key in data:
@@ -72,7 +74,9 @@ def cmd_metadata(self):
             self.die("error deleting metadata: %s", str(e))
     else:
         # Get (or unspecified)
-        keys = self.args.get or None
+        keys = None
+        if self.args.get:
+            keys = map(nilmdb.utils.unicode.decode, self.args.get)
         try:
             data = self.client.stream_get_metadata(self.args.path, keys)
         except nilmdb.client.ClientError as e:
@@ -81,4 +85,6 @@ def cmd_metadata(self):
             # Print nonexistant keys as having empty value
             if value is None:
                 value = ""
-            printf("%s=%s\n", key, value)
+            printf("%s=%s\n",
+                   nilmdb.utils.unicode.encode(key),
+                   nilmdb.utils.unicode.encode(value))

nilmdb/cmdline/remove.py

@@ -1,17 +1,19 @@
 from nilmdb.utils.printf import *
 import nilmdb.client
+import fnmatch
 
 def setup(self, sub):
     cmd = sub.add_parser("remove", help="Remove data",
                          description="""
                          Remove all data from a specified time range within a
-                         stream.
+                         stream.  If multiple streams or wildcards are provided,
+                         the same time range is removed from all streams.
                          """)
     cmd.set_defaults(handler = cmd_remove)
 
     group = cmd.add_argument_group("Data selection")
-    group.add_argument("path",
-                       help="Path of stream, e.g. /foo/bar",
+    group.add_argument("path", nargs='+',
+                       help="Path of stream, e.g. /foo/bar/*",
                        ).completer = self.complete.path
     group.add_argument("-s", "--start", required=True,
                        metavar="TIME", type=self.arg_time,
@@ -23,18 +25,31 @@ def setup(self, sub):
                        ).completer = self.complete.time
 
     group = cmd.add_argument_group("Output format")
+    group.add_argument("-q", "--quiet", action="store_true",
+                       help="Don't display names when removing "
+                       "from multiple paths")
     group.add_argument("-c", "--count", action="store_true",
                        help="Output number of data points removed")
     return cmd
 
 def cmd_remove(self):
+    streams = [ s[0] for s in self.client.stream_list() ]
+    paths = []
+    for path in self.args.path:
+        new = fnmatch.filter(streams, path)
+        if not new:
+            self.die("error: no stream matched path: %s", path)
+        paths.extend(new)
+
     try:
-        count = self.client.stream_remove(self.args.path,
-                                          self.args.start, self.args.end)
+        for path in paths:
+            if not self.args.quiet and len(paths) > 1:
+                printf("Removing from %s\n", path)
+            count = self.client.stream_remove(path,
+                                              self.args.start, self.args.end)
+            if self.args.count:
+                printf("%d\n", count);
     except nilmdb.client.ClientError as e:
         self.die("error removing data: %s", str(e))
 
-    if self.args.count:
-        printf("%d\n", count)
-
     return 0

nilmdb/fsck/__init__.py

@@ -0,0 +1,5 @@
+"""nilmdb.fsck"""
+
+from __future__ import absolute_import
+
+from nilmdb.fsck.fsck import Fsck

nilmdb/fsck/fsck.py

@@ -0,0 +1,458 @@
+# -*- coding: utf-8 -*-
+
+"""Check database consistency, with some ability to fix problems.
+This should be able to fix cases where a database gets corrupted due
+to unexpected system shutdown, and detect other cases that may cause
+NilmDB to return errors when trying to manipulate the database."""
+
+import nilmdb.utils
+import nilmdb.server
+import nilmdb.client.numpyclient
+from nilmdb.utils.interval import IntervalError
+from nilmdb.server.interval import Interval, IntervalSet
+from nilmdb.utils.printf import *
+from nilmdb.utils.time import timestamp_to_string
+
+from collections import defaultdict
+import sqlite3
+import os
+import sys
+import progressbar
+import re
+import time
+import shutil
+import cPickle as pickle
+import numpy
+
+class FsckError(Exception):
+    def __init__(self, msg = "", *args):
+        if args:
+            msg = sprintf(msg, *args)
+        Exception.__init__(self, msg)
+class FixableFsckError(FsckError):
+    def __init__(self, msg = "", *args):
+        if args:
+            msg = sprintf(msg, *args)
+        FsckError.__init__(self, "%s\nThis may be fixable with \"--fix\".", msg)
+class RetryFsck(FsckError):
+    pass
+
+def log(format, *args):
+    printf(format, *args)
+
+def err(format, *args):
+    fprintf(sys.stderr, format, *args)
+
+# Decorator that retries a function if it returns a specific value
+def retry_if_raised(exc, message = None, max_retries = 100):
+    def f1(func):
+        def f2(*args, **kwargs):
+            for n in range(max_retries):
+                try:
+                    return func(*args, **kwargs)
+                except exc as e:
+                    if message:
+                        log("%s\n\n", message)
+            raise Exception("Max number of retries (%d) exceeded; giving up")
+        return f2
+    return f1
+
+class Progress(object):
+    def __init__(self, maxval):
+        self.bar = progressbar.ProgressBar(
+            maxval = maxval,
+            widgets = [ progressbar.Percentage(), ' ',
+                        progressbar.Bar(), ' ',
+                        progressbar.ETA() ])
+        if self.bar.term_width == 0:
+            self.bar.term_width = 75
+    def __enter__(self):
+        self.bar.start()
+        self.last_update = 0
+        return self
+    def __exit__(self, exc_type, exc_value, traceback):
+        if exc_type is None:
+            self.bar.finish()
+        else:
+            printf("\n")
+    def update(self, val):
+        self.bar.update(val)
+
+class Fsck(object):
+
+    def __init__(self, path, fix = False):
+        self.basepath = path
+        self.sqlpath = os.path.join(path, "data.sql")
+        self.bulkpath = os.path.join(path, "data")
+        self.bulklock = os.path.join(path, "data.lock")
+        self.fix = fix
+
+    ### Main checks
+
+    @retry_if_raised(RetryFsck, "Something was fixed: restarting fsck")
+    def check(self, skip_data = False):
+        self.bulk = None
+        self.sql = None
+        try:
+            self.check_paths()
+            self.check_sql()
+            self.check_streams()
+            self.check_intervals()
+            if skip_data:
+                log("skipped data check\n")
+            else:
+                self.check_data()
+        finally:
+            if self.bulk:
+                self.bulk.close()
+            if self.sql:
+                self.sql.commit()
+                self.sql.close()
+        log("ok\n")
+
+    ### Check basic path structure
+
+    def check_paths(self):
+        log("checking paths\n")
+        if self.bulk:
+            self.bulk.close()
+        if not os.path.isfile(self.sqlpath):
+            raise FsckError("SQL database missing (%s)", self.sqlpath)
+        if not os.path.isdir(self.bulkpath):
+            raise FsckError("Bulk data directory missing (%s)", self.bulkpath)
+        with open(self.bulklock, "w") as lockfile:
+            if not nilmdb.utils.lock.exclusive_lock(lockfile):
+                raise FsckError('Database already locked by another process\n'
+                                'Make sure all other processes that might be '
+                                'using the database are stopped.\n'
+                                'Restarting apache will cause it to unlock '
+                                'the db until a request is received.')
+            # unlocked immediately
+        self.bulk = nilmdb.server.bulkdata.BulkData(self.basepath)
+
+    ### Check SQL database health
+
+    def check_sql(self):
+        log("checking sqlite database\n")
+
+        self.sql = sqlite3.connect(self.sqlpath)
+        with self.sql:
+            cur = self.sql.cursor()
+            ver = cur.execute("PRAGMA user_version").fetchone()[0]
+            good = max(nilmdb.server.nilmdb._sql_schema_updates.keys())
+            if ver != good:
+                raise FsckError("database version %d too old, should be %d",
+                                ver, good)
+            self.stream_path = {}
+            self.stream_layout = {}
+            log("  loading paths\n")
+            result = cur.execute("SELECT id, path, layout FROM streams")
+            for r in result:
+                if r[0] in self.stream_path:
+                    raise FsckError("duplicated ID %d in stream IDs", r[0])
+                self.stream_path[r[0]] = r[1]
+                self.stream_layout[r[0]] = r[2]
+
+            log("  loading intervals\n")
+            self.stream_interval = defaultdict(list)
+            result = cur.execute("SELECT stream_id, start_time, end_time, "
+                                 "start_pos, end_pos FROM ranges "
+                                 "ORDER BY start_time")
+            for r in result:
+                if r[0] not in self.stream_path:
+                    raise FsckError("interval ID %d not in streams", k)
+                self.stream_interval[r[0]].append((r[1], r[2], r[3], r[4]))
+
+            log("  loading metadata\n")
+            self.stream_meta = defaultdict(dict)
+            result = cur.execute("SELECT stream_id, key, value FROM metadata")
+            for r in result:
+                if r[0] not in self.stream_path:
+                    raise FsckError("metadata ID %d not in streams", k)
+                if r[1] in self.stream_meta[r[0]]:
+                    raise FsckError("duplicate metadata key '%s' for stream %d",
+                                    r[1], r[0])
+                self.stream_meta[r[0]][r[1]] = r[2]
+
+    ### Check streams and basic interval overlap
+
+    def check_streams(self):
+        ids = self.stream_path.keys()
+        log("checking %s streams\n", "{:,d}".format(len(ids)))
+        with Progress(len(ids)) as pbar:
+            for i, sid in enumerate(ids):
+                pbar.update(i)
+                path = self.stream_path[sid]
+
+                # unique path, valid layout
+                if self.stream_path.values().count(path) != 1:
+                    raise FsckError("duplicated path %s", path)
+                layout = self.stream_layout[sid].split('_')[0]
+                if layout not in ('int8', 'int16', 'int32', 'int64',
+                                  'uint8', 'uint16', 'uint32', 'uint64',
+                                  'float32', 'float64'):
+                    raise FsckError("bad layout %s for %s", layout, path)
+                count = int(self.stream_layout[sid].split('_')[1])
+                if count < 1 or count > 1024:
+                    raise FsckError("bad count %d for %s", count, path)
+
+                # must exist in bulkdata
+                bulk = self.bulkpath + path
+                if not os.path.isdir(bulk):
+                    raise FsckError("%s: missing bulkdata dir", path)
+                if not nilmdb.server.bulkdata.Table.exists(bulk):
+                    raise FsckError("%s: bad bulkdata table", path)
+
+                # intervals don't overlap.  Abuse IntervalSet to check
+                # for intervals in file positions, too.
+                timeiset = IntervalSet()
+                posiset = IntervalSet()
+                for (stime, etime, spos, epos) in self.stream_interval[sid]:
+                    new = Interval(stime, etime)
+                    try:
+                        timeiset += new
+                    except IntervalError:
+                        raise FsckError("%s: overlap in intervals:\n"
+                                        "set: %s\nnew: %s",
+                                        path, str(timeiset), str(new))
+                    if spos != epos:
+                        new = Interval(spos, epos)
+                        try:
+                            posiset += new
+                        except IntervalError:
+                            raise FsckError("%s: overlap in file offsets:\n"
+                                            "set: %s\nnew: %s",
+                                            path, str(posiset), str(new))
+
+                # check bulkdata
+                self.check_bulkdata(sid, path, bulk)
+
+                # Check that we can open bulkdata
+                try:
+                    tab = None
+                    try:
+                        tab = nilmdb.server.bulkdata.Table(bulk)
+                    except Exception as e:
+                        raise FsckError("%s: can't open bulkdata: %s",
+                                        path, str(e))
+                finally:
+                    if tab:
+                        tab.close()
+
+    ### Check that bulkdata is good enough to be opened
+
+    @retry_if_raised(RetryFsck)
+    def check_bulkdata(self, sid, path, bulk):
+        with open(os.path.join(bulk, "_format"), "rb") as f:
+            fmt = pickle.load(f)
+        if fmt["version"] != 3:
+            raise FsckError("%s: bad or unsupported bulkdata version %d",
+                            path, fmt["version"])
+        row_per_file = int(fmt["rows_per_file"])
+        files_per_dir = int(fmt["files_per_dir"])
+        layout = fmt["layout"]
+        if layout != self.stream_layout[sid]:
+            raise FsckError("%s: layout mismatch %s != %s", path,
+                            layout, self.stream_layout[sid])
+
+        # Every file should have a size that's the multiple of the row size
+        rkt = nilmdb.server.rocket.Rocket(layout, None)
+        row_size = rkt.binary_size
+        rkt.close()
+
+        # Find all directories
+        regex = re.compile("^[0-9a-f]{4,}$")
+        subdirs = sorted(filter(regex.search, os.listdir(bulk)),
+                         key = lambda x: int(x, 16), reverse = True)
+        for subdir in subdirs:
+            # Find all files in that dir
+            subpath = os.path.join(bulk, subdir)
+            files = filter(regex.search, os.listdir(subpath))
+            if not files:
+                self.fix_empty_subdir(subpath)
+                raise RetryFsck
+            # Verify that their size is a multiple of the row size
+            for filename in files:
+                filepath = os.path.join(subpath, filename)
+                offset = os.path.getsize(filepath)
+                if offset % row_size:
+                    self.fix_bad_filesize(path, filepath, offset, row_size)
+
+    def fix_empty_subdir(self, subpath):
+        msg = sprintf("bulkdata path %s is missing data files", subpath)
+        if not self.fix:
+            raise FixableFsckError(msg)
+        # Try to fix it by just deleting whatever is present,
+        # as long as it's only ".removed" files.
+        err("\n%s\n", msg)
+        for fn in os.listdir(subpath):
+            if not fn.endswith(".removed"):
+                raise FsckError("can't fix automatically: please manually "
+                                "remove the file %s and try again",
+                                os.path.join(subpath, fn))
+        # Remove the whole thing
+        err("Removing empty subpath\n")
+        shutil.rmtree(subpath)
+        raise RetryFsck
+
+    def fix_bad_filesize(self, path, filepath, offset, row_size):
+        extra = offset % row_size
+        msg = sprintf("%s: size of file %s (%d) is not a multiple" +
+                      " of row size (%d): %d extra bytes present",
+                      path, filepath, offset, row_size, extra)
+        if not self.fix:
+            raise FixableFsckError(msg)
+        # Try to fix it by just truncating the file
+        err("\n%s\n", msg)
+        newsize = offset - extra
+        err("Truncating file to %d bytes and retrying\n", newsize)
+        with open(filepath, "r+b") as f:
+            f.truncate(newsize)
+            raise RetryFsck
+
+    ### Check interval endpoints
+
+    def check_intervals(self):
+        total_ints = sum(len(x) for x in self.stream_interval.values())
+        log("checking %s intervals\n", "{:,d}".format(total_ints))
+        done = 0
+        with Progress(total_ints) as pbar:
+            for sid in self.stream_interval:
+                try:
+                    bulk = self.bulkpath + self.stream_path[sid]
+                    tab = nilmdb.server.bulkdata.Table(bulk)
+                    def update(x):
+                        pbar.update(done + x)
+                    ints = self.stream_interval[sid]
+                    done += self.check_table_intervals(sid, ints, tab, update)
+                finally:
+                    tab.close()
+
+    def check_table_intervals(self, sid, ints, tab, update):
+        # look in the table to make sure we can pick out the interval's
+        # endpoints
+        path = self.stream_path[sid]
+        tab.file_open.cache_remove_all()
+        for (i, intv) in enumerate(ints):
+            update(i)
+            (stime, etime, spos, epos) = intv
+            if spos == epos and spos >= 0 and spos <= tab.nrows:
+                continue
+            try:
+                srow = tab[spos]
+                erow = tab[epos-1]
+            except Exception as e:
+                self.fix_bad_interval(sid, intv, tab, str(e))
+                raise RetryFsck
+        return len(ints)
+
+    def fix_bad_interval(self, sid, intv, tab, msg):
+        path = self.stream_path[sid]
+        msg = sprintf("%s: interval %s error accessing rows: %s",
+                      path, str(intv), str(msg))
+        if not self.fix:
+            raise FixableFsckError(msg)
+        err("\n%s\n", msg)
+
+        (stime, etime, spos, epos) = intv
+        # If it's just that the end pos is more than the number of rows
+        # in the table, lower end pos and truncate interval time too.
+        if spos < tab.nrows and epos >= tab.nrows:
+            err("end position is past endrows, but it can be truncated\n")
+            err("old end: time %d, pos %d\n", etime, epos)
+            new_epos = tab.nrows
+            new_etime = tab[new_epos-1] + 1
+            err("new end: time %d, pos %d\n", new_etime, new_epos)
+            if stime < new_etime:
+                # Change it in SQL
+                with self.sql:
+                    cur = self.sql.cursor()
+                    cur.execute("UPDATE ranges SET end_time=?, end_pos=? "
+                                "WHERE stream_id=? AND start_time=? AND "
+                                "end_time=? AND start_pos=? AND end_pos=?",
+                                (new_etime, new_epos, sid, stime, etime,
+                                 spos, epos))
+                    if cur.rowcount != 1:
+                        raise FsckError("failed to fix SQL database")
+                raise RetryFsck
+            err("actually it can't be truncated; times are bad too")
+
+        # Otherwise, the only hope is to delete the interval entirely.
+        err("*** Deleting the entire interval from SQL.\n")
+        err("This may leave stale data on disk.  To fix that, copy all\n")
+        err("data from this stream to a new stream, then remove all data\n")
+        err("from and destroy %s.\n")
+        with self.sql:
+            cur = self.sql.cursor()
+            cur.execute("DELETE FROM ranges WHERE "
+                        "stream_id=? AND start_time=? AND "
+                        "end_time=? AND start_pos=? AND end_pos=?",
+                        (sid, stime, etime, spos, epos))
+            if cur.rowcount != 1:
+                raise FsckError("failed to remove interval")
+        raise RetryFsck
+
+    ### Check data in each interval
+
+    def check_data(self):
+        total_rows = sum(sum((y[3] - y[2]) for y in x)
+                         for x in self.stream_interval.values())
+        log("checking %s rows of data\n", "{:,d}".format(total_rows))
+        done = 0
+        with Progress(total_rows) as pbar:
+            for sid in self.stream_interval:
+                try:
+                    bulk = self.bulkpath + self.stream_path[sid]
+                    tab = nilmdb.server.bulkdata.Table(bulk)
+                    def update(x):
+                        pbar.update(done + x)
+                    ints = self.stream_interval[sid]
+                    done += self.check_table_data(sid, ints, tab, update)
+                finally:
+                    tab.close()
+
+    def check_table_data(self, sid, ints, tab, update):
+        # Pull out all of the interval's data and verify that it's
+        # monotonic.
+        maxrows = 100000
+        path = self.stream_path[sid]
+        layout = self.stream_layout[sid]
+        dtype = nilmdb.client.numpyclient.layout_to_dtype(layout)
+        tab.file_open.cache_remove_all()
+        done = 0
+        for intv in ints:
+            last_ts = None
+            (stime, etime, spos, epos) = intv
+            if spos == epos:
+                continue
+            for start in xrange(*slice(spos, epos, maxrows).indices(epos)):
+                stop = min(start + maxrows, epos)
+                count = stop - start
+                # Get raw data, convert to NumPy arary
+                try:
+                    raw = tab.get_data(start, stop, binary = True)
+                    data = numpy.fromstring(raw, dtype)
+                except Exception as e:
+                    raise FsckError("%s: failed to grab rows %d through %d: %s",
+                                    path, start, stop, repr(e))
+
+                # Verify that timestamps are monotonic
+                if (numpy.diff(data['timestamp']) <= 0).any():
+                    raise FsckError("%s: non-monotonic timestamp(s) in rows "
+                                    "%d through %d", path, start, stop)
+                first_ts = data['timestamp'][0]
+                if last_ts is not None and first_ts <= last_ts:
+                    raise FsckError("%s: first interval timestamp %d is not "
+                                    "greater than the previous last interval "
+                                    "timestamp %d, at row %d",
+                                    path, first_ts, last_ts, start)
+                last_ts = data['timestamp'][-1]
+
+                # These are probably fixable, by removing the offending
+                # intervals.  But I'm not going to bother implementing
+                # that yet.
+
+                # Done
+                done += count
+                update(done)
+        return done

nilmdb/scripts/nilmdb_fsck.py

@@ -0,0 +1,26 @@
+#!/usr/bin/python
+
+import nilmdb.fsck
+import argparse
+import os
+import sys
+
+def main():
+    """Main entry point for the 'nilmdb-fsck' command line script"""
+
+    parser = argparse.ArgumentParser(
+        description = 'Check database consistency',
+        formatter_class = argparse.ArgumentDefaultsHelpFormatter,
+        version = nilmdb.__version__)
+    parser.add_argument("-f", "--fix", action="store_true",
+                        default=False, help = 'Fix errors when possible '
+                        '(which may involve removing data)')
+    parser.add_argument("-n", "--no-data", action="store_true",
+                        default=False, help = 'Skip the slow full-data check')
+    parser.add_argument('database', help = 'Database directory')
+    args = parser.parse_args()
+
+    nilmdb.fsck.Fsck(args.database, args.fix).check(skip_data = args.no_data)
+
+if __name__ == "__main__":
+    main()

nilmdb/scripts/nilmdb_server.py

@@ -10,10 +10,8 @@ def main():
 
     parser = argparse.ArgumentParser(
         description = 'Run the NilmDB server',
-        formatter_class = argparse.ArgumentDefaultsHelpFormatter)
-
-    parser.add_argument("-V", "--version", action="version",
-                        version = nilmdb.__version__)
+        formatter_class = argparse.ArgumentDefaultsHelpFormatter,
+        version = nilmdb.__version__)
 
     group = parser.add_argument_group("Standard options")
     group.add_argument('-a', '--address',

nilmdb/server/bulkdata.py

@@ -19,8 +19,8 @@ from . import rocket
 
 # Up to 256 open file descriptors at any given time.
 # These variables are global so they can be used in the decorator arguments.
-table_cache_size = 16
-fd_cache_size = 16
+table_cache_size = 32
+fd_cache_size = 8
 
 @nilmdb.utils.must_close(wrap_verify = False)
 class BulkData(object):
@@ -330,7 +330,8 @@ class Table(object):
 
         # Find the last directory.  We sort and loop through all of them,
         # starting with the numerically greatest, because the dirs could be
-        # empty if something was deleted.
+        # empty if something was deleted but the directory was unexpectedly
+        # not deleted.
         subdirs = sorted(filter(regex.search, os.listdir(self.root)),
                          key = lambda x: int(x, 16), reverse = True)
 

nilmdb/server/nilmdb.py

@@ -176,7 +176,7 @@ class NilmDB(object):
             raise NilmDBError("start must precede end")
         return (start, end)
 
-    @nilmdb.utils.lru_cache(size = 16)
+    @nilmdb.utils.lru_cache(size = 64)
     def _get_intervals(self, stream_id):
         """
         Return a mutable IntervalSet corresponding to the given stream ID.
@@ -675,6 +675,7 @@ class NilmDB(object):
 
             # Count how many were removed
             removed += row_end - row_start
+            remaining -= row_end - row_start
 
             if restart is not None:
                 break

nilmdb/server/rocket.c

@@ -5,6 +5,9 @@
 #include <ctype.h>
 #include <stdint.h>
 
+#define __STDC_FORMAT_MACROS
+#include <inttypes.h>
+
 /* Values missing from stdint.h */
 #define UINT8_MIN 0
 #define UINT16_MIN 0
@@ -19,16 +22,9 @@
 
 typedef int64_t timestamp_t;
 
-/* This code probably needs to be double-checked for the case where
-   sizeof(long) != 8, so enforce that here with something that will
-   fail at build time.  We assume that the python integer type can
-   hold an int64_t. */
-const static char __long_ok[1 - 2*!(sizeof(int64_t) ==
-				    sizeof(long int))] = { 0 };
-
 /* Somewhat arbitrary, just so we can use fixed sizes for strings
    etc. */
-static const int MAX_LAYOUT_COUNT = 128;
+static const int MAX_LAYOUT_COUNT = 1024;
 
 /* Error object and constants */
 static PyObject *ParseError;
@@ -58,7 +54,7 @@ static PyObject *raise_str(int line, int col, int code, const char *string)
 static PyObject *raise_int(int line, int col, int code, int64_t num)
 {
 	PyObject *o;
-	o = Py_BuildValue("(iiil)", line, col, code, num);
+	o = Py_BuildValue("(iiiL)", line, col, code, (long long)num);
 	if (o != NULL) {
 		PyErr_SetObject(ParseError, o);
 		Py_DECREF(o);
@@ -249,11 +245,11 @@ static PyObject *Rocket_get_file_size(Rocket *self)
 /****
  * Append from string
  */
-static inline long int strtol10(const char *nptr, char **endptr) {
-	return strtol(nptr, endptr, 10);
+static inline long int strtoll10(const char *nptr, char **endptr) {
+	return strtoll(nptr, endptr, 10);
 }
-static inline long int strtoul10(const char *nptr, char **endptr) {
-	return strtoul(nptr, endptr, 10);
+static inline long int strtoull10(const char *nptr, char **endptr) {
+	return strtoull(nptr, endptr, 10);
 }
 
 /* .append_string(count, data, offset, linenum, start, end, last_timestamp) */
@@ -264,6 +260,7 @@ static PyObject *Rocket_append_string(Rocket *self, PyObject *args)
 	int offset;
 	const char *linestart;
 	int linenum;
+        long long ll1, ll2, ll3;
 	timestamp_t start;
 	timestamp_t end;
 	timestamp_t last_timestamp;
@@ -280,10 +277,13 @@ static PyObject *Rocket_append_string(Rocket *self, PyObject *args)
 	   but we need the null termination for strto*.  If we had
 	   strnto* that took a length, we could use t# and not require
 	   a copy. */
-	if (!PyArg_ParseTuple(args, "isiilll:append_string", &count,
+	if (!PyArg_ParseTuple(args, "isiiLLL:append_string", &count,
 			      &data, &offset, &linenum,
-			      &start, &end, &last_timestamp))
+			      &ll1, &ll2, &ll3))
 		return NULL;
+        start = ll1;
+        end = ll2;
+        last_timestamp = ll3;
 
 	/* Skip spaces, but don't skip over a newline. */
 #define SKIP_BLANK(buf) do {			\
@@ -372,14 +372,14 @@ static PyObject *Rocket_append_string(Rocket *self, PyObject *args)
 				goto extra_data_on_line;		\
 			break
 
-			CS(INT8,   strtol10,  t64.i, t8.i,  t8.u,         , 1);
-			CS(UINT8,  strtoul10, t64.u, t8.u,  t8.u,         , 1);
-			CS(INT16,  strtol10,  t64.i, t16.i, t16.u, le16toh, 2);
-			CS(UINT16, strtoul10, t64.u, t16.u, t16.u, le16toh, 2);
-			CS(INT32,  strtol10,  t64.i, t32.i, t32.u, le32toh, 4);
-			CS(UINT32, strtoul10, t64.u, t32.u, t32.u, le32toh, 4);
-			CS(INT64,  strtol10,  t64.i, t64.i, t64.u, le64toh, 8);
-			CS(UINT64, strtoul10, t64.u, t64.u, t64.u, le64toh, 8);
+			CS(INT8,   strtoll10,  t64.i, t8.i,  t8.u,         , 1);
+			CS(UINT8,  strtoull10, t64.u, t8.u,  t8.u,         , 1);
+			CS(INT16,  strtoll10,  t64.i, t16.i, t16.u, le16toh, 2);
+			CS(UINT16, strtoull10, t64.u, t16.u, t16.u, le16toh, 2);
+			CS(INT32,  strtoll10,  t64.i, t32.i, t32.u, le32toh, 4);
+			CS(UINT32, strtoull10, t64.u, t32.u, t32.u, le32toh, 4);
+			CS(INT64,  strtoll10,  t64.i, t64.i, t64.u, le64toh, 8);
+			CS(UINT64, strtoull10, t64.u, t64.u, t64.u, le64toh, 8);
 			CS(FLOAT32, strtod,   t64.d, t32.f, t32.u, le32toh, 4);
 			CS(FLOAT64, strtod,   t64.d, t64.d, t64.u, le64toh, 8);
 #undef CS
@@ -397,7 +397,8 @@ static PyObject *Rocket_append_string(Rocket *self, PyObject *args)
 	/* Build return value and return */
 	offset = buf - data;
 	PyObject *o;
-	o = Py_BuildValue("(iili)", written, offset, last_timestamp, linenum);
+	o = Py_BuildValue("(iiLi)", written, offset,
+                          (long long)last_timestamp, linenum);
 	return o;
 err:
 	PyErr_SetFromErrno(PyExc_OSError);
@@ -431,14 +432,18 @@ static PyObject *Rocket_append_binary(Rocket *self, PyObject *args)
         int data_len;
         int linenum;
 	int offset;
+        long long ll1, ll2, ll3;
 	timestamp_t start;
 	timestamp_t end;
 	timestamp_t last_timestamp;
 
-	if (!PyArg_ParseTuple(args, "it#iilll:append_binary",
+	if (!PyArg_ParseTuple(args, "it#iiLLL:append_binary",
                               &count, &data, &data_len, &offset,
-                              &linenum, &start, &end, &last_timestamp))
+                              &linenum, &ll1, &ll2, &ll3))
 		return NULL;
+        start = ll1;
+        end = ll2;
+        last_timestamp = ll3;
 
         /* Advance to offset */
         if (offset > data_len)
@@ -468,7 +473,7 @@ static PyObject *Rocket_append_binary(Rocket *self, PyObject *args)
         }
 
         /* Write binary data */
-        if (fwrite(data, data_len, 1, self->file) != 1) {
+        if (fwrite(data, self->binary_size, rows, self->file) != rows) {
                 PyErr_SetFromErrno(PyExc_OSError);
                 return NULL;
         }
@@ -476,8 +481,8 @@ static PyObject *Rocket_append_binary(Rocket *self, PyObject *args)
 
 	/* Build return value and return */
 	PyObject *o;
-	o = Py_BuildValue("(iili)", rows, offset + rows * self->binary_size,
-                          last_timestamp, linenum);
+	o = Py_BuildValue("(iiLi)", rows, offset + rows * self->binary_size,
+                          (long long)last_timestamp, linenum);
 	return o;
 }
 
@@ -534,7 +539,7 @@ static PyObject *Rocket_extract_string(Rocket *self, PyObject *args)
 		if (fread(&t64.u, 8, 1, self->file) != 1)
 			goto err;
 		t64.u = le64toh(t64.u);
-		ret = sprintf(&str[len], "%ld", t64.i);
+		ret = sprintf(&str[len], "%" PRId64, t64.i);
 		if (ret <= 0)
 			goto err;
 		len += ret;
@@ -556,14 +561,14 @@ static PyObject *Rocket_extract_string(Rocket *self, PyObject *args)
 				len += ret;				\
 			}						\
 			break
-			CASE(INT8,   "%hhd",   t8.i,  t8.u,         , 1);
-			CASE(UINT8,  "%hhu",   t8.u,  t8.u,         , 1);
-			CASE(INT16,  "%hd",    t16.i, t16.u, le16toh, 2);
-			CASE(UINT16, "%hu",    t16.u, t16.u, le16toh, 2);
-			CASE(INT32,  "%d",     t32.i, t32.u, le32toh, 4);
-			CASE(UINT32, "%u",     t32.u, t32.u, le32toh, 4);
-			CASE(INT64,  "%ld",    t64.i, t64.u, le64toh, 8);
-			CASE(UINT64, "%lu",    t64.u, t64.u, le64toh, 8);
+			CASE(INT8,   "%" PRId8,  t8.i,  t8.u,         , 1);
+			CASE(UINT8,  "%" PRIu8,  t8.u,  t8.u,         , 1);
+			CASE(INT16,  "%" PRId16, t16.i, t16.u, le16toh, 2);
+			CASE(UINT16, "%" PRIu16, t16.u, t16.u, le16toh, 2);
+			CASE(INT32,  "%" PRId32, t32.i, t32.u, le32toh, 4);
+			CASE(UINT32, "%" PRIu32, t32.u, t32.u, le32toh, 4);
+			CASE(INT64,  "%" PRId64, t64.i, t64.u, le64toh, 8);
+			CASE(UINT64, "%" PRIu64, t64.u, t64.u, le64toh, 8);
 			/* These next two are a bit debatable.  floats
 			   are 6-9 significant figures, so we print 7.
 			   Doubles are 15-19, so we print 17.  This is
@@ -653,7 +658,7 @@ static PyObject *Rocket_extract_timestamp(Rocket *self, PyObject *args)
 
 	/* Convert and return */
 	t64.u = le64toh(t64.u);
-	return Py_BuildValue("l", t64.i);
+	return Py_BuildValue("L", (long long)t64.i);
 }
 
 /****

nilmdb/server/server.py

@@ -17,126 +17,26 @@ import decorator
 import psutil
 import traceback
 
+from nilmdb.server.serverutil import (
+    chunked_response,
+    response_type,
+    workaround_cp_bug_1200,
+    exception_to_httperror,
+    CORS_allow,
+    json_to_request_params,
+    json_error_page,
+    cherrypy_start,
+    cherrypy_stop,
+    bool_param,
+    )
+
+# Add CORS_allow tool
+cherrypy.tools.CORS_allow = cherrypy.Tool('on_start_resource', CORS_allow)
+
 class NilmApp(object):
     def __init__(self, db):
         self.db = db
 
-# Decorators
-def chunked_response(func):
-    """Decorator to enable chunked responses."""
-    # Set this to False to get better tracebacks from some requests
-    # (/stream/extract, /stream/intervals).
-    func._cp_config = { 'response.stream': True }
-    return func
-
-def response_type(content_type):
-    """Return a decorator-generating function that sets the
-    response type to the specified string."""
-    def wrapper(func, *args, **kwargs):
-        cherrypy.response.headers['Content-Type'] = content_type
-        return func(*args, **kwargs)
-    return decorator.decorator(wrapper)
-
-@decorator.decorator
-def workaround_cp_bug_1200(func, *args, **kwargs): # pragma: no cover
-    """Decorator to work around CherryPy bug #1200 in a response
-    generator.
-
-    Even if chunked responses are disabled, LookupError or
-    UnicodeError exceptions may still be swallowed by CherryPy due to
-    bug #1200.  This throws them as generic Exceptions instead so that
-    they make it through.
-    """
-    exc_info = None
-    try:
-        for val in func(*args, **kwargs):
-            yield val
-    except (LookupError, UnicodeError):
-        # Re-raise it, but maintain the original traceback
-        exc_info = sys.exc_info()
-        new_exc = Exception(exc_info[0].__name__ + ": " + str(exc_info[1]))
-        raise new_exc, None, exc_info[2]
-    finally:
-        del exc_info
-
-def exception_to_httperror(*expected):
-    """Return a decorator-generating function that catches expected
-    errors and throws a HTTPError describing it instead.
-
-        @exception_to_httperror(NilmDBError, ValueError)
-        def foo():
-            pass
-    """
-    def wrapper(func, *args, **kwargs):
-        exc_info = None
-        try:
-            return func(*args, **kwargs)
-        except expected:
-            # Re-raise it, but maintain the original traceback
-            exc_info = sys.exc_info()
-            new_exc = cherrypy.HTTPError("400 Bad Request", str(exc_info[1]))
-            raise new_exc, None, exc_info[2]
-        finally:
-            del exc_info
-    # We need to preserve the function's argspecs for CherryPy to
-    # handle argument errors correctly.  Decorator.decorator takes
-    # care of that.
-    return decorator.decorator(wrapper)
-
-# Custom CherryPy tools
-
-def CORS_allow(methods):
-    """This does several things:
-
-    Handles CORS preflight requests.
-    Adds Allow: header to all requests.
-    Raise 405 if request.method not in method.
-
-    It is similar to cherrypy.tools.allow, with the CORS stuff added.
-    """
-    request = cherrypy.request.headers
-    response = cherrypy.response.headers
-
-    if not isinstance(methods, (tuple, list)): # pragma: no cover
-        methods = [ methods ]
-    methods = [ m.upper() for m in methods if m ]
-    if not methods: # pragma: no cover
-        methods = [ 'GET', 'HEAD' ]
-    elif 'GET' in methods and 'HEAD' not in methods: # pragma: no cover
-        methods.append('HEAD')
-    response['Allow'] = ', '.join(methods)
-
-    # Allow all origins
-    if 'Origin' in request:
-        response['Access-Control-Allow-Origin'] = request['Origin']
-
-    # If it's a CORS request, send response.
-    request_method = request.get("Access-Control-Request-Method", None)
-    request_headers = request.get("Access-Control-Request-Headers", None)
-    if (cherrypy.request.method == "OPTIONS" and
-        request_method and request_headers):
-        response['Access-Control-Allow-Headers'] = request_headers
-        response['Access-Control-Allow-Methods'] = ', '.join(methods)
-        # Try to stop further processing and return a 200 OK
-        cherrypy.response.status = "200 OK"
-        cherrypy.response.body = ""
-        cherrypy.request.handler = lambda: ""
-        return
-
-    # Reject methods that were not explicitly allowed
-    if cherrypy.request.method not in methods:
-        raise cherrypy.HTTPError(405)
-
-cherrypy.tools.CORS_allow = cherrypy.Tool('on_start_resource', CORS_allow)
-
-# Helper for json_in tool to process JSON data into normal request
-# parameters.
-def json_to_request_params(body):
-    cherrypy.lib.jsontools.json_processor(body)
-    if not isinstance(cherrypy.request.json, dict):
-        raise cherrypy.HTTPError(415)
-    cherrypy.request.params.update(cherrypy.request.json)
-
 # CherryPy apps
 class Root(NilmApp):
     """Root application for NILM database"""
@@ -147,7 +47,10 @@ class Root(NilmApp):
     # /
     @cherrypy.expose
     def index(self):
-        raise cherrypy.NotFound()
+        cherrypy.response.headers['Content-Type'] = 'text/plain'
+        msg = sprintf("This is NilmDB version %s, running on host %s.\n",
+                      nilmdb.__version__, socket.getfqdn())
+        return msg
 
     # /favicon.ico
     @cherrypy.expose
@@ -167,9 +70,13 @@ class Root(NilmApp):
         """Return a dictionary with the database path,
         size of the database in bytes, and free disk space in bytes"""
         path = self.db.get_basepath()
+        usage = psutil.disk_usage(path)
+        dbsize = nilmdb.utils.du(path)
         return { "path": path,
-                 "size": nilmdb.utils.du(path),
-                 "free": psutil.disk_usage(path).free }
+                 "size": dbsize,
+                 "other": max(usage.used - dbsize, 0),
+                 "reserved": max(usage.total - usage.used - usage.free, 0),
+                 "free": usage.free }
 
 class Stream(NilmApp):
     """Stream-specific operations"""
@@ -177,10 +84,18 @@ class Stream(NilmApp):
     # Helpers
     def _get_times(self, start_param, end_param):
         (start, end) = (None, None)
-        if start_param is not None:
-            start = string_to_timestamp(start_param)
-        if end_param is not None:
-            end = string_to_timestamp(end_param)
+        try:
+            if start_param is not None:
+                start = string_to_timestamp(start_param)
+        except Exception:
+            raise cherrypy.HTTPError("400 Bad Request", sprintf(
+                "invalid start (%s): must be a numeric timestamp", start_param))
+        try:
+            if end_param is not None:
+                end = string_to_timestamp(end_param)
+        except Exception:
+            raise cherrypy.HTTPError("400 Bad Request", sprintf(
+                "invalid end (%s): must be a numeric timestamp", end_param))
         if start is not None and end is not None:
             if start >= end:
                 raise cherrypy.HTTPError(
@@ -199,10 +114,10 @@ class Stream(NilmApp):
         layout parameter, just list streams that match the given path
         or layout.
 
-        If extent is not given, returns a list of lists containing
-        the path and layout: [ path, layout ]
+        If extended is missing or zero, returns a list of lists
+        containing the path and layout: [ path, layout ]
 
-        If extended is provided, returns a list of lists containing
+        If extended is true, returns a list of lists containing
         extended info: [ path, layout, extent_min, extent_max,
         total_rows, total_seconds ].  More data may be added.
         """
@@ -315,6 +230,8 @@ class Stream(NilmApp):
         little-endian and matches the database types (including an
         int64 timestamp).
         """
+        binary = bool_param(binary)
+
         # Important that we always read the input before throwing any
         # errors, to keep lengths happy for persistent connections.
         # Note that CherryPy 3.2.2 has a bug where this fails for GET
@@ -347,24 +264,34 @@ class Stream(NilmApp):
     # /stream/remove?path=/newton/prep&start=1234567890.0&end=1234567899.0
     @cherrypy.expose
     @cherrypy.tools.json_in()
-    @cherrypy.tools.json_out()
-    @exception_to_httperror(NilmDBError)
     @cherrypy.tools.CORS_allow(methods = ["POST"])
+    @chunked_response
+    @response_type("application/x-json-stream")
     def remove(self, path, start = None, end = None):
         """
         Remove data from the backend database.  Removes all data in
-        the interval [start, end).  Returns the number of data points
-        removed.
+        the interval [start, end).
+
+        Returns the number of data points removed.  Since this is a potentially
+        long-running operation, multiple numbers may be returned as the
+        data gets removed from the backend database.  The total number of
+        points removed is the sum of all of these numbers.
         """
         (start, end) = self._get_times(start, end)
-        total_removed = 0
-        while True:
-            (removed, restart) = self.db.stream_remove(path, start, end)
-            total_removed += removed
-            if restart is None:
-                break
-            start = restart
-        return total_removed
+
+        if len(self.db.stream_list(path = path)) != 1:
+            raise cherrypy.HTTPError("404", "No such stream: " + path)
+
+        @workaround_cp_bug_1200
+        def content(start, end):
+            # Note: disable chunked responses to see tracebacks from here.
+            while True:
+                (removed, restart) = self.db.stream_remove(path, start, end)
+                yield json.dumps(removed) + "\r\n"
+                if restart is None:
+                    break
+                start = restart
+        return content(start, end)
 
     # /stream/intervals?path=/newton/prep
     # /stream/intervals?path=/newton/prep&start=1234567890.0&end=1234567899.0
@@ -429,6 +356,10 @@ class Stream(NilmApp):
         little-endian and matches the database types (including an
         int64 timestamp).
         """
+        binary = bool_param(binary)
+        markup = bool_param(markup)
+        count = bool_param(count)
+
         (start, end) = self._get_times(start, end)
 
         # Check path and get layout
@@ -556,70 +487,14 @@ class Server(object):
 
     def json_error_page(self, status, message, traceback, version):
         """Return a custom error page in JSON so the client can parse it"""
-        errordata = { "status" : status,
-                      "message" : message,
-                      "traceback" : traceback }
-        # Don't send a traceback if the error was 400-499 (client's fault)
-        try:
-            code = int(status.split()[0])
-            if not self.force_traceback:
-                if code >= 400 and code <= 499:
-                    errordata["traceback"] = ""
-        except Exception: # pragma: no cover
-            pass
-        # Override the response type, which was previously set to text/html
-        cherrypy.serving.response.headers['Content-Type'] = (
-            "application/json;charset=utf-8" )
-        # Undo the HTML escaping that cherrypy's get_error_page function applies
-        # (cherrypy issue 1135)
-        for k, v in errordata.iteritems():
-            v = v.replace("&lt;","<")
-            v = v.replace("&gt;",">")
-            v = v.replace("&amp;","&")
-            errordata[k] = v
-        return json.dumps(errordata, separators=(',',':'))
+        return json_error_page(status, message, traceback, version,
+                               self.force_traceback)
 
     def start(self, blocking = False, event = None):
-
-        if not self.embedded: # pragma: no cover
-            # Handle signals nicely
-            if hasattr(cherrypy.engine, "signal_handler"):
-                cherrypy.engine.signal_handler.subscribe()
-            if hasattr(cherrypy.engine, "console_control_handler"):
-                cherrypy.engine.console_control_handler.subscribe()
-
-        # Cherrypy stupidly calls os._exit(70) when it can't bind the
-        # port.  At least try to print a reasonable error and continue
-        # in this case, rather than just dying silently (as we would
-        # otherwise do in embedded mode)
-        real_exit = os._exit
-        def fake_exit(code): # pragma: no cover
-            if code == os.EX_SOFTWARE:
-                fprintf(sys.stderr, "error: CherryPy called os._exit!\n")
-            else:
-                real_exit(code)
-        os._exit = fake_exit
-        cherrypy.engine.start()
-        os._exit = real_exit
-
-        # Signal that the engine has started successfully
-        if event is not None:
-            event.set()
-
-        if blocking:
-            try:
-                cherrypy.engine.wait(cherrypy.engine.states.EXITING,
-                                     interval = 0.1, channel = 'main')
-            except (KeyboardInterrupt, IOError): # pragma: no cover
-                cherrypy.engine.log('Keyboard Interrupt: shutting down bus')
-                cherrypy.engine.exit()
-            except SystemExit: # pragma: no cover
-                cherrypy.engine.log('SystemExit raised: shutting down bus')
-                cherrypy.engine.exit()
-                raise
+        cherrypy_start(blocking, event, self.embedded)
 
     def stop(self):
-        cherrypy.engine.exit()
+        cherrypy_stop()
 
 # Use a single global nilmdb.server.NilmDB and nilmdb.server.Server
 # instance since the database can only be opened once.  For this to

nilmdb/server/serverutil.py

@@ -0,0 +1,214 @@
+"""Miscellaneous decorators and other helpers for running a CherryPy
+server"""
+
+import cherrypy
+import sys
+import os
+import decorator
+import simplejson as json
+
+# Helper to parse parameters into booleans
+def bool_param(s):
+    """Return a bool indicating whether parameter 's' was True or False,
+    supporting a few different types for 's'."""
+    try:
+        ss = s.lower()
+        if ss in [ "0", "false", "f", "no", "n" ]:
+            return False
+        if ss in [ "1", "true", "t", "yes", "y" ]:
+            return True
+    except Exception:
+        return bool(s)
+    raise cherrypy.HTTPError("400 Bad Request",
+                             "can't parse parameter: " + ss)
+
+# Decorators
+def chunked_response(func):
+    """Decorator to enable chunked responses."""
+    # Set this to False to get better tracebacks from some requests
+    # (/stream/extract, /stream/intervals).
+    func._cp_config = { 'response.stream': True }
+    return func
+
+def response_type(content_type):
+    """Return a decorator-generating function that sets the
+    response type to the specified string."""
+    def wrapper(func, *args, **kwargs):
+        cherrypy.response.headers['Content-Type'] = content_type
+        return func(*args, **kwargs)
+    return decorator.decorator(wrapper)
+
+@decorator.decorator
+def workaround_cp_bug_1200(func, *args, **kwargs): # pragma: no cover
+    """Decorator to work around CherryPy bug #1200 in a response
+    generator.
+
+    Even if chunked responses are disabled, LookupError or
+    UnicodeError exceptions may still be swallowed by CherryPy due to
+    bug #1200.  This throws them as generic Exceptions instead so that
+    they make it through.
+    """
+    exc_info = None
+    try:
+        for val in func(*args, **kwargs):
+            yield val
+    except (LookupError, UnicodeError):
+        # Re-raise it, but maintain the original traceback
+        exc_info = sys.exc_info()
+        new_exc = Exception(exc_info[0].__name__ + ": " + str(exc_info[1]))
+        raise new_exc, None, exc_info[2]
+    finally:
+        del exc_info
+
+def exception_to_httperror(*expected):
+    """Return a decorator-generating function that catches expected
+    errors and throws a HTTPError describing it instead.
+
+        @exception_to_httperror(NilmDBError, ValueError)
+        def foo():
+            pass
+    """
+    def wrapper(func, *args, **kwargs):
+        exc_info = None
+        try:
+            return func(*args, **kwargs)
+        except expected:
+            # Re-raise it, but maintain the original traceback
+            exc_info = sys.exc_info()
+            new_exc = cherrypy.HTTPError("400 Bad Request", str(exc_info[1]))
+            raise new_exc, None, exc_info[2]
+        finally:
+            del exc_info
+    # We need to preserve the function's argspecs for CherryPy to
+    # handle argument errors correctly.  Decorator.decorator takes
+    # care of that.
+    return decorator.decorator(wrapper)
+
+# Custom CherryPy tools
+
+def CORS_allow(methods):
+    """This does several things:
+
+    Handles CORS preflight requests.
+    Adds Allow: header to all requests.
+    Raise 405 if request.method not in method.
+
+    It is similar to cherrypy.tools.allow, with the CORS stuff added.
+
+    Add this to CherryPy with:
+    cherrypy.tools.CORS_allow = cherrypy.Tool('on_start_resource', CORS_allow)
+    """
+    request = cherrypy.request.headers
+    response = cherrypy.response.headers
+
+    if not isinstance(methods, (tuple, list)): # pragma: no cover
+        methods = [ methods ]
+    methods = [ m.upper() for m in methods if m ]
+    if not methods: # pragma: no cover
+        methods = [ 'GET', 'HEAD' ]
+    elif 'GET' in methods and 'HEAD' not in methods: # pragma: no cover
+        methods.append('HEAD')
+    response['Allow'] = ', '.join(methods)
+
+    # Allow all origins
+    if 'Origin' in request:
+        response['Access-Control-Allow-Origin'] = request['Origin']
+
+    # If it's a CORS request, send response.
+    request_method = request.get("Access-Control-Request-Method", None)
+    request_headers = request.get("Access-Control-Request-Headers", None)
+    if (cherrypy.request.method == "OPTIONS" and
+        request_method and request_headers):
+        response['Access-Control-Allow-Headers'] = request_headers
+        response['Access-Control-Allow-Methods'] = ', '.join(methods)
+        # Try to stop further processing and return a 200 OK
+        cherrypy.response.status = "200 OK"
+        cherrypy.response.body = ""
+        cherrypy.request.handler = lambda: ""
+        return
+
+    # Reject methods that were not explicitly allowed
+    if cherrypy.request.method not in methods:
+        raise cherrypy.HTTPError(405)
+
+
+# Helper for json_in tool to process JSON data into normal request
+# parameters.
+def json_to_request_params(body):
+    cherrypy.lib.jsontools.json_processor(body)
+    if not isinstance(cherrypy.request.json, dict):
+        raise cherrypy.HTTPError(415)
+    cherrypy.request.params.update(cherrypy.request.json)
+
+# Used as an "error_page.default" handler
+def json_error_page(status, message, traceback, version,
+                    force_traceback = False):
+    """Return a custom error page in JSON so the client can parse it"""
+    errordata = { "status" : status,
+                  "message" : message,
+                  "traceback" : traceback }
+    # Don't send a traceback if the error was 400-499 (client's fault)
+    try:
+        code = int(status.split()[0])
+        if not force_traceback:
+            if code >= 400 and code <= 499:
+                errordata["traceback"] = ""
+    except Exception: # pragma: no cover
+        pass
+    # Override the response type, which was previously set to text/html
+    cherrypy.serving.response.headers['Content-Type'] = (
+        "application/json;charset=utf-8" )
+    # Undo the HTML escaping that cherrypy's get_error_page function applies
+    # (cherrypy issue 1135)
+    for k, v in errordata.iteritems():
+        v = v.replace("&lt;","<")
+        v = v.replace("&gt;",">")
+        v = v.replace("&amp;","&")
+        errordata[k] = v
+    return json.dumps(errordata, separators=(',',':'))
+
+# Start/stop CherryPy standalone server
+def cherrypy_start(blocking = False, event = False, embedded = False):
+    """Start the CherryPy server, handling errors and signals
+    somewhat gracefully."""
+
+    if not embedded: # pragma: no cover
+        # Handle signals nicely
+        if hasattr(cherrypy.engine, "signal_handler"):
+            cherrypy.engine.signal_handler.subscribe()
+        if hasattr(cherrypy.engine, "console_control_handler"):
+            cherrypy.engine.console_control_handler.subscribe()
+
+    # Cherrypy stupidly calls os._exit(70) when it can't bind the
+    # port.  At least try to print a reasonable error and continue
+    # in this case, rather than just dying silently (as we would
+    # otherwise do in embedded mode)
+    real_exit = os._exit
+    def fake_exit(code): # pragma: no cover
+        if code == os.EX_SOFTWARE:
+            fprintf(sys.stderr, "error: CherryPy called os._exit!\n")
+        else:
+            real_exit(code)
+    os._exit = fake_exit
+    cherrypy.engine.start()
+    os._exit = real_exit
+
+    # Signal that the engine has started successfully
+    if event is not None:
+        event.set()
+
+    if blocking:
+        try:
+            cherrypy.engine.wait(cherrypy.engine.states.EXITING,
+                                 interval = 0.1, channel = 'main')
+        except (KeyboardInterrupt, IOError): # pragma: no cover
+            cherrypy.engine.log('Keyboard Interrupt: shutting down bus')
+            cherrypy.engine.exit()
+        except SystemExit: # pragma: no cover
+            cherrypy.engine.log('SystemExit raised: shutting down bus')
+            cherrypy.engine.exit()
+            raise
+
+# Stop CherryPy server
+def cherrypy_stop():
+    cherrypy.engine.exit()

nilmdb/utils/__init__.py

@@ -13,3 +13,5 @@ import nilmdb.utils.time
 import nilmdb.utils.iterator
 import nilmdb.utils.interval
 import nilmdb.utils.lock
+import nilmdb.utils.sort
+import nilmdb.utils.unicode

nilmdb/utils/diskusage.py

@@ -1,4 +1,5 @@
 import os
+import errno
 from math import log
 
 def human_size(num):
@@ -16,10 +17,18 @@ def human_size(num):
         return '1 byte'
 
 def du(path):
-    """Like du -sb, returns total size of path in bytes."""
-    size = os.path.getsize(path)
-    if os.path.isdir(path):
-        for thisfile in os.listdir(path):
-            filepath = os.path.join(path, thisfile)
-            size += du(filepath)
-    return size
+    """Like du -sb, returns total size of path in bytes.  Ignore
+    errors that might occur if we encounter broken symlinks or
+    files in the process of being removed."""
+    try:
+        st = os.stat(path)
+        size = st.st_blocks * 512
+        if os.path.isdir(path):
+            for thisfile in os.listdir(path):
+                filepath = os.path.join(path, thisfile)
+                size += du(filepath)
+        return size
+    except OSError as e: # pragma: no cover
+        if e.errno != errno.ENOENT:
+            raise
+        return 0

nilmdb/utils/interval.py

@@ -1,5 +1,6 @@
 """Interval.  Like nilmdb.server.interval, but re-implemented here
-in plain Python so clients have easier access to it.
+in plain Python so clients have easier access to it, and with a few
+helper functions.
 
 Intervals are half-open, ie. they include data points with timestamps
 [start, end)
@@ -34,6 +35,10 @@ class Interval:
         return ("[" + nilmdb.utils.time.timestamp_to_string(self.start) +
                 " -> " + nilmdb.utils.time.timestamp_to_string(self.end) + ")")
 
+    def human_string(self):
+        return ("[ " + nilmdb.utils.time.timestamp_to_human(self.start) +
+                " -> " + nilmdb.utils.time.timestamp_to_human(self.end) + " ]")
+
     def __cmp__(self, other):
         """Compare two intervals.  If non-equal, order by start then end"""
         return cmp(self.start, other.start) or cmp(self.end, other.end)
@@ -53,18 +58,11 @@ class Interval:
             raise IntervalError("not a subset")
         return Interval(start, end)
 
-def set_difference(a, b):
-    """
-    Compute the difference (a \\ b) between the intervals in 'a' and
-    the intervals in 'b'; i.e., the ranges that are present in 'self'
-    but not 'other'.
-
-    'a' and 'b' must both be iterables.
-
-    Returns a generator that yields each interval in turn.
-    Output intervals are built as subsets of the intervals in the
-    first argument (a).
-    """
+def _interval_math_helper(a, b, op, subset = True):
+    """Helper for set_difference, intersection functions,
+    to compute interval subsets based on a math operator on ranges
+    present in A and B.  Subsets are computed from A, or new intervals
+    are generated if subset = False."""
     # Iterate through all starts and ends in sorted order.  Add a
     # tag to the iterator so that we can figure out which one they
     # were, after sorting.
@@ -79,28 +77,71 @@ def set_difference(a, b):
     # At each point, evaluate which type of end it is, to determine
     # how to build up the output intervals.
     a_interval = None
-    b_interval = None
+    in_a = False
+    in_b = False
     out_start = None
     for (ts, k, i) in nilmdb.utils.iterator.imerge(a_iter, b_iter):
         if k == 0:
-            # start a interval
             a_interval = i
-            if b_interval is None:
-                out_start = ts
+            in_a = True
         elif k == 1:
-            # start b interval
-            b_interval = i
-            if out_start is not None and out_start != ts:
-                yield a_interval.subset(out_start, ts)
-            out_start = None
+            in_b = True
         elif k == 2:
-            # end a interval
-            if out_start is not None and out_start != ts:
-                yield a_interval.subset(out_start, ts)
-            out_start = None
-            a_interval = None
+            in_a = False
         elif k == 3:
-            # end b interval
-            b_interval = None
-            if a_interval:
-                out_start = ts
+            in_b = False
+        include = op(in_a, in_b)
+        if include and out_start is None:
+            out_start = ts
+        elif not include:
+            if out_start is not None and out_start != ts:
+                if subset:
+                    yield a_interval.subset(out_start, ts)
+                else:
+                    yield Interval(out_start, ts)
+            out_start = None
+
+def set_difference(a, b):
+    """
+    Compute the difference (a \\ b) between the intervals in 'a' and
+    the intervals in 'b'; i.e., the ranges that are present in 'self'
+    but not 'other'.
+
+    'a' and 'b' must both be iterables.
+
+    Returns a generator that yields each interval in turn.
+    Output intervals are built as subsets of the intervals in the
+    first argument (a).
+    """
+    return _interval_math_helper(a, b, (lambda a, b: a and not b))
+
+def intersection(a, b):
+    """
+    Compute the intersection between the intervals in 'a' and the
+    intervals in 'b'; i.e., the ranges that are present in both 'a'
+    and 'b'.
+
+    'a' and 'b' must both be iterables.
+
+    Returns a generator that yields each interval in turn.
+    Output intervals are built as subsets of the intervals in the
+    first argument (a).
+    """
+    return _interval_math_helper(a, b, (lambda a, b: a and b))
+
+def optimize(it):
+    """
+    Given an iterable 'it' with intervals, optimize them by joining
+    together intervals that are adjacent in time, and return a generator
+    that yields the new intervals.
+    """
+    saved_int = None
+    for interval in it:
+        if saved_int is not None:
+            if saved_int.end == interval.start:
+                interval.start = saved_int.start
+            else:
+                yield saved_int
+        saved_int = interval
+    if saved_int is not None:
+        yield saved_int

nilmdb/utils/serializer.py

@@ -91,6 +91,20 @@ def serializer_proxy(obj_or_type):
             r = SerializerCallProxy(self.__call_queue, attr, self)
             return r
 
+        # For an interable object, on __iter__(), save the object's
+        # iterator and return this proxy.  On next(), call the object's
+        # iterator through this proxy.
+        def __iter__(self):
+            attr = getattr(self.__object, "__iter__")
+            self.__iter = SerializerCallProxy(self.__call_queue, attr, self)()
+            return self
+        def next(self):
+            return SerializerCallProxy(self.__call_queue,
+                                       self.__iter.next, self)()
+
+        def __getitem__(self, key):
+            return self.__getattr__("__getitem__")(key)
+
         def __call__(self, *args, **kwargs):
             """Call this to instantiate the type, if a type was passed
             to serializer_proxy.  Otherwise, pass the call through."""

nilmdb/utils/sort.py

@@ -0,0 +1,18 @@
+import re
+
+def sort_human(items, key = None):
+    """Human-friendly sort (/stream/2 before /stream/10)"""
+    def to_num(val):
+        try:
+            return int(val)
+        except Exception:
+            return val
+
+    def human_key(text):
+        if key:
+            text = key(text)
+        # Break into character and numeric chunks.
+        chunks = re.split(r'([0-9]+)', text)
+        return [ to_num(c) for c in chunks ]
+
+    return sorted(items, key = human_key)

nilmdb/utils/time.py

@@ -60,7 +60,7 @@ def rate_to_period(hz, cycles = 1):
 def parse_time(toparse):
     """
     Parse a free-form time string and return a nilmdb timestamp
-    (integer seconds since epoch).  If the string doesn't contain a
+    (integer microseconds since epoch).  If the string doesn't contain a
     timestamp, the current local timezone is assumed (e.g. from the TZ
     env var).
     """

nilmdb/utils/unicode.py

@@ -0,0 +1,29 @@
+import sys
+
+if sys.version_info[0] >= 3: # pragma: no cover (future Python3 compat)
+    text_type = str
+else:
+    text_type = unicode
+
+def encode(u):
+    """Try to encode something from Unicode to a string using the
+    default encoding.  If it fails, try encoding as UTF-8."""
+    if not isinstance(u, text_type):
+        return u
+    try:
+        return u.encode()
+    except UnicodeEncodeError:
+        return u.encode("utf-8")
+
+def decode(s):
+    """Try to decode someting from string to Unicode using the
+    default encoding.  If it fails, try decoding as UTF-8."""
+    if isinstance(s, text_type):
+        return s
+    try:
+        return s.decode()
+    except UnicodeDecodeError:
+        try:
+            return s.decode("utf-8")
+        except UnicodeDecodeError:
+            return s # best we can do

setup.py

@@ -118,6 +118,7 @@ setup(name='nilmdb',
                            'pytz',
                            'psutil >= 0.3.0',
                            'requests >= 1.1.0, < 2.0.0',
+                           'progressbar >= 2.2',
                            ],
       packages = [ 'nilmdb',
                    'nilmdb.utils',
@@ -126,11 +127,13 @@ setup(name='nilmdb',
                    'nilmdb.client',
                    'nilmdb.cmdline',
                    'nilmdb.scripts',
+                   'nilmdb.fsck',
                    ],
       entry_points = {
           'console_scripts': [
               'nilmtool = nilmdb.scripts.nilmtool:main',
               'nilmdb-server = nilmdb.scripts.nilmdb_server:main',
+              'nilmdb-fsck = nilmdb.scripts.nilmdb_fsck:main',
               ],
           },
       ext_modules = ext_modules,

tests/data/timestamped

@@ -0,0 +1,8 @@
+-10000000000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+-100000000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+-100000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+-1000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+1 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+1000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+1000000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+1000000000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  

tests/test_client.py

@@ -105,16 +105,19 @@ class TestClient(object):
                 client.http.post("/stream/list")
         client = nilmdb.client.Client(url = testurl)
 
-        # Create three streams
+        # Create four streams
         client.stream_create("/newton/prep", "float32_8")
         client.stream_create("/newton/raw", "uint16_6")
-        client.stream_create("/newton/zzz/rawnotch", "uint16_9")
+        client.stream_create("/newton/zzz/rawnotch2", "uint16_9")
+        client.stream_create("/newton/zzz/rawnotch11", "uint16_9")
 
-        # Verify we got 3 streams
+        # Verify we got 4 streams in the right order
         eq_(client.stream_list(), [ ["/newton/prep", "float32_8"],
                                     ["/newton/raw", "uint16_6"],
-                                    ["/newton/zzz/rawnotch", "uint16_9"]
+                                    ["/newton/zzz/rawnotch2", "uint16_9"],
+                                    ["/newton/zzz/rawnotch11", "uint16_9"]
                                     ])
+
         # Match just one type or one path
         eq_(client.stream_list(layout="uint16_6"),
             [ ["/newton/raw", "uint16_6"] ])
@@ -239,6 +242,19 @@ class TestClient(object):
         in_("400 Bad Request", str(e.exception))
         in_("start must precede end", str(e.exception))
 
+        # Invalid times in HTTP request
+        with assert_raises(ClientError) as e:
+            client.http.put("stream/insert", "", { "path": "/newton/prep",
+                                                   "start": "asdf", "end": 0 })
+        in_("400 Bad Request", str(e.exception))
+        in_("invalid start", str(e.exception))
+
+        with assert_raises(ClientError) as e:
+            client.http.put("stream/insert", "", { "path": "/newton/prep",
+                                                   "start": 0, "end": "asdf" })
+        in_("400 Bad Request", str(e.exception))
+        in_("invalid end", str(e.exception))
+
         # Good content type
         with assert_raises(ClientError) as e:
             client.http.put("stream/insert", "",
@@ -327,6 +343,10 @@ class TestClient(object):
                      2525.169921875, 8350.83984375, 3724.699951171875,
                      1355.3399658203125, 2039.0))
 
+        # Just get some coverage
+        with assert_raises(ClientError) as e:
+            client.http.post("/stream/remove", { "path": "/none" })
+
         client.close()
 
     def test_client_06_generators(self):
@@ -347,10 +367,6 @@ class TestClient(object):
         with assert_raises(ServerError) as e:
             client.http.get_gen("http://nosuchurl.example.com./").next()
 
-        # Trigger a curl error in generator
-        with assert_raises(ServerError) as e:
-            client.http.get_gen("http://nosuchurl.example.com./").next()
-
         # Check 404 for missing streams
         for function in [ client.stream_intervals, client.stream_extract ]:
             with assert_raises(ClientError) as e:
@@ -389,27 +405,38 @@ class TestClient(object):
                                  headers())
 
         # Extract
-        x = http.get("stream/extract",
-                            { "path": "/newton/prep",
-                              "start": "123",
-                              "end": "124" })
+        x = http.get("stream/extract", { "path": "/newton/prep",
+                                         "start": "123", "end": "124" })
         if "transfer-encoding: chunked" not in headers():
             warnings.warn("Non-chunked HTTP response for /stream/extract")
         if "content-type: text/plain;charset=utf-8" not in headers():
             raise AssertionError("/stream/extract is not text/plain:\n" +
                                  headers())
 
-        x = http.get("stream/extract",
-                            { "path": "/newton/prep",
-                              "start": "123",
-                              "end": "124",
-                              "binary": "1" })
+        x = http.get("stream/extract", { "path": "/newton/prep",
+                                         "start": "123", "end": "124",
+                                         "binary": "1" })
         if "transfer-encoding: chunked" not in headers():
             warnings.warn("Non-chunked HTTP response for /stream/extract")
         if "content-type: application/octet-stream" not in headers():
             raise AssertionError("/stream/extract is not binary:\n" +
                                  headers())
 
+        # Make sure a binary of "0" is really off
+        x = http.get("stream/extract", { "path": "/newton/prep",
+                                         "start": "123", "end": "124",
+                                         "binary": "0" })
+        if "content-type: application/octet-stream" in headers():
+                    raise AssertionError("/stream/extract is not text:\n" +
+                                         headers())
+
+        # Invalid parameters
+        with assert_raises(ClientError) as e:
+            x = http.get("stream/extract", { "path": "/newton/prep",
+                                             "start": "123", "end": "124",
+                                             "binary": "asdfasfd" })
+        in_("can't parse parameter", str(e.exception))
+
         client.close()
 
     def test_client_08_unicode(self):
@@ -613,8 +640,12 @@ class TestClient(object):
         with client.stream_insert_context("/empty/test", end = 950):
             pass
 
+        # Equal start and end is OK as long as there's no data
+        with client.stream_insert_context("/empty/test", start=9, end=9):
+            pass
+
         # Try various things that might cause problems
-        with client.stream_insert_context("/empty/test", 1000, 1050):
+        with client.stream_insert_context("/empty/test", 1000, 1050) as ctx:
             ctx.finalize() # inserts [1000, 1050]
             ctx.finalize() # nothing
             ctx.finalize() # nothing

tests/test_cmdline.py

@@ -59,8 +59,7 @@ class TestCmdline(object):
 
     def run(self, arg_string, infile=None, outfile=None):
         """Run a cmdline client with the specified argument string,
-        passing the given input.  Returns a tuple with the output and
-        exit code"""
+        passing the given input.  Save the output and exit code."""
         # printf("TZ=UTC ./nilmtool.py %s\n", arg_string)
         os.environ['NILMDB_URL'] = "http://localhost:32180/"
         class stdio_wrapper:
@@ -88,7 +87,7 @@ class TestCmdline(object):
                 sys.exit(0)
             except SystemExit as e:
                 exitcode = e.code
-        captured = outfile.getvalue()
+        captured = nilmdb.utils.unicode.decode(outfile.getvalue())
         self.captured = captured
         self.exitcode = exitcode
 
@@ -160,6 +159,12 @@ class TestCmdline(object):
         self.ok("--help")
         self.contain("usage:")
 
+        # help
+        self.ok("--version")
+        ver = self.captured
+        self.ok("list --version")
+        eq_(self.captured, ver)
+
         # fail for no args
         self.fail("")
 
@@ -245,8 +250,10 @@ class TestCmdline(object):
         self.contain("Client version: " + nilmdb.__version__)
         self.contain("Server version: " + test_server.version)
         self.contain("Server database path")
-        self.contain("Server database size")
-        self.contain("Server database free space")
+        self.contain("Server disk space used by NilmDB")
+        self.contain("Server disk space used by other")
+        self.contain("Server disk space reserved")
+        self.contain("Server disk space free")
 
     def test_04_createlist(self):
         # Basic stream tests, like those in test_client.
@@ -300,38 +307,19 @@ class TestCmdline(object):
 
         # Verify we got those 3 streams and they're returned in
         # alphabetical order.
-        self.ok("list")
+        self.ok("list -l")
         self.match("/newton/prep float32_8\n"
                    "/newton/raw uint16_6\n"
                    "/newton/zzz/rawnotch uint16_9\n")
 
         # Match just one type or one path.  Also check
         # that --path is optional
-        self.ok("list --path /newton/raw")
-        self.match("/newton/raw uint16_6\n")
-
-        self.ok("list /newton/raw")
-        self.match("/newton/raw uint16_6\n")
-
-        self.fail("list -p /newton/raw /newton/raw")
-        self.contain("too many paths")
-
-        self.ok("list --layout uint16_6")
+        self.ok("list --layout /newton/raw")
         self.match("/newton/raw uint16_6\n")
 
         # Wildcard matches
-        self.ok("list --layout uint16*")
-        self.match("/newton/raw uint16_6\n"
-                   "/newton/zzz/rawnotch uint16_9\n")
-
-        self.ok("list --path *zzz* --layout uint16*")
-        self.match("/newton/zzz/rawnotch uint16_9\n")
-
-        self.ok("list *zzz* --layout uint16*")
-        self.match("/newton/zzz/rawnotch uint16_9\n")
-
-        self.ok("list --path *zzz* --layout float32*")
-        self.match("")
+        self.ok("list *zzz*")
+        self.match("/newton/zzz/rawnotch\n")
 
         # reversed range
         self.fail("list /newton/prep --start 2020-01-01 --end 2000-01-01")
@@ -492,33 +480,40 @@ class TestCmdline(object):
         # bad start time
         self.fail("insert -t -r 120 --start 'whatever' /newton/prep /dev/null")
 
+        # Test negative times
+        self.ok("insert --start @-10000000000 --end @1000000001 /newton/prep"
+                " tests/data/timestamped")
+        self.ok("extract -c /newton/prep --start min --end @1000000001")
+        self.match("8\n")
+        self.ok("remove /newton/prep --start min --end @1000000001")
+
     def test_07_detail_extended(self):
         # Just count the number of lines, it's probably fine
         self.ok("list --detail")
         lines_(self.captured, 8)
 
-        self.ok("list --detail --path *prep")
+        self.ok("list --detail *prep")
         lines_(self.captured, 4)
 
-        self.ok("list --detail --path *prep --start='23 Mar 2012 10:02'")
+        self.ok("list --detail *prep --start='23 Mar 2012 10:02'")
         lines_(self.captured, 3)
 
-        self.ok("list --detail --path *prep --start='23 Mar 2012 10:05'")
+        self.ok("list --detail *prep --start='23 Mar 2012 10:05'")
         lines_(self.captured, 2)
 
-        self.ok("list --detail --path *prep --start='23 Mar 2012 10:05:15'")
+        self.ok("list --detail *prep --start='23 Mar 2012 10:05:15'")
         lines_(self.captured, 2)
         self.contain("10:05:15.000")
 
-        self.ok("list --detail --path *prep --start='23 Mar 2012 10:05:15.50'")
+        self.ok("list --detail *prep --start='23 Mar 2012 10:05:15.50'")
         lines_(self.captured, 2)
         self.contain("10:05:15.500")
 
-        self.ok("list --detail --path *prep --start='23 Mar 2012 19:05:15.50'")
+        self.ok("list --detail *prep --start='23 Mar 2012 19:05:15.50'")
         lines_(self.captured, 2)
         self.contain("no intervals")
 
-        self.ok("list --detail --path *prep --start='23 Mar 2012 10:05:15.50'"
+        self.ok("list --detail *prep --start='23 Mar 2012 10:05:15.50'"
                 + " --end='23 Mar 2012 10:05:15.51'")
         lines_(self.captured, 2)
         self.contain("10:05:15.500")
@@ -527,15 +522,15 @@ class TestCmdline(object):
         lines_(self.captured, 8)
 
         # Verify the "raw timestamp" output
-        self.ok("list --detail --path *prep --timestamp-raw "
+        self.ok("list --detail *prep --timestamp-raw "
                 "--start='23 Mar 2012 10:05:15.50'")
         lines_(self.captured, 2)
         self.contain("[ 1332497115500000 -> 1332497160000000 ]")
 
         # bad time
-        self.fail("list --detail --path *prep -T --start='9332497115.612'")
+        self.fail("list --detail *prep -T --start='9332497115.612'")
         # good time
-        self.ok("list --detail --path *prep -T --start='1332497115.612'")
+        self.ok("list --detail *prep -T --start='1332497115.612'")
         lines_(self.captured, 2)
         self.contain("[ 1332497115612000 -> 1332497160000000 ]")
 
@@ -620,6 +615,14 @@ class TestCmdline(object):
         self.ok("extract -c /newton/prep --start 2000-01-01 --end 2020-01-01")
         self.match("43200\n")
 
+        # test binary mode
+        self.fail("extract -c -B /newton/prep -s min -e max")
+        self.contain("binary cannot be combined")
+        self.fail("extract -m -B /newton/prep -s min -e max")
+        self.contain("binary cannot be combined")
+        self.ok("extract -B /newton/prep -s min -e max")
+        eq_(len(self.captured), 43200 * (8 + 8*4))
+
         # markup for 3 intervals, plus extra markup lines whenever we had
         # a "restart" from the nilmdb.stream_extract function
         self.ok("extract -m /newton/prep --start 2000-01-01 --end 2020-01-01")
@@ -639,7 +642,7 @@ class TestCmdline(object):
 
         # Try nonexistent stream
         self.fail("remove /no/such/foo --start 2000-01-01 --end 2020-01-01")
-        self.contain("No stream at path")
+        self.contain("no stream matched path")
 
         # empty or backward ranges return errors
         self.fail("remove /newton/prep --start 2020-01-01 --end 2000-01-01")
@@ -667,9 +670,14 @@ class TestCmdline(object):
                 "--start '23 Mar 2022 20:00:30' " +
                 "--end '23 Mar 2022 20:00:31'")
         self.match("0\n")
+        self.ok("remove -c /newton/prep /newton/pre* " +
+                "--start '23 Mar 2022 20:00:30' " +
+                "--end '23 Mar 2022 20:00:31'")
+        self.match("Removing from /newton/prep\n0\n" +
+                   "Removing from /newton/prep\n0\n")
 
         # Make sure we have the data we expect
-        self.ok("list --detail /newton/prep")
+        self.ok("list -l --detail /newton/prep")
         self.match("/newton/prep float32_8\n" +
                    "  [ Fri, 23 Mar 2012 10:00:00.000000 +0000"
                    " -> Fri, 23 Mar 2012 10:01:59.991668 +0000 ]\n"
@@ -704,7 +712,7 @@ class TestCmdline(object):
         self.match("24000\n")
 
         # See the missing chunks in list output
-        self.ok("list --detail /newton/prep")
+        self.ok("list --layout --detail /newton/prep")
         self.match("/newton/prep float32_8\n" +
                    "  [ Fri, 23 Mar 2012 10:00:00.000000 +0000"
                    " -> Fri, 23 Mar 2012 10:00:05.000000 +0000 ]\n"
@@ -718,7 +726,7 @@ class TestCmdline(object):
         # Remove all data, verify it's missing
         self.ok("remove /newton/prep --start 2000-01-01 --end 2020-01-01")
         self.match("")  # no count requested this time
-        self.ok("list --detail /newton/prep")
+        self.ok("list -l --detail /newton/prep")
         self.match("/newton/prep float32_8\n" +
                    "  (no intervals)\n")
 
@@ -736,16 +744,16 @@ class TestCmdline(object):
         self.contain("too few arguments")
 
         self.fail("destroy /no/such/stream")
-        self.contain("No stream at path")
+        self.contain("no stream matched path")
 
         self.fail("destroy -R /no/such/stream")
-        self.contain("No stream at path")
+        self.contain("no stream matched path")
 
         self.fail("destroy asdfasdf")
-        self.contain("No stream at path")
+        self.contain("no stream matched path")
 
         # From previous tests, we have:
-        self.ok("list")
+        self.ok("list -l")
         self.match("/newton/prep float32_8\n"
                    "/newton/raw uint16_6\n"
                    "/newton/zzz/rawnotch uint16_9\n")
@@ -761,13 +769,13 @@ class TestCmdline(object):
         lines_(self.captured, 7)
 
         # Destroy for real
-        self.ok("destroy -R /newton/prep")
-        self.ok("list")
+        self.ok("destroy -R /n*/prep")
+        self.ok("list -l")
         self.match("/newton/raw uint16_6\n"
                    "/newton/zzz/rawnotch uint16_9\n")
 
         self.ok("destroy /newton/zzz/rawnotch")
-        self.ok("list")
+        self.ok("list -l")
         self.match("/newton/raw uint16_6\n")
 
         self.ok("destroy /newton/raw")
@@ -786,18 +794,17 @@ class TestCmdline(object):
             self.ok("list")
             self.contain(path)
             # Make sure it was created empty
-            self.ok("list --detail --path " + path)
+            self.ok("list --detail " + path)
             self.contain("(no intervals)")
 
     def test_12_unicode(self):
         # Unicode paths.
         self.ok("destroy /newton/asdf/qwer")
-        self.ok("destroy /newton/prep")
-        self.ok("destroy /newton/raw")
+        self.ok("destroy /newton/prep /newton/raw")
         self.ok("destroy /newton/zzz")
 
         self.ok(u"create /düsseldorf/raw uint16_6")
-        self.ok("list --detail")
+        self.ok("list -l --detail")
         self.contain(u"/düsseldorf/raw uint16_6")
         self.contain("(no intervals)")
 
@@ -883,7 +890,7 @@ class TestCmdline(object):
         du_before = nilmdb.utils.diskusage.du(testdb)
 
         # Make sure we have the data we expect
-        self.ok("list --detail")
+        self.ok("list -l --detail")
         self.match("/newton/prep float32_8\n" +
                    "  [ Fri, 23 Mar 2012 10:00:00.000000 +0000"
                    " -> Fri, 23 Mar 2012 10:01:59.991668 +0000 ]\n"
@@ -919,7 +926,7 @@ class TestCmdline(object):
         self.match("3600\n")
 
         # See the missing chunks in list output
-        self.ok("list --detail")
+        self.ok("list -l --detail")
         self.match("/newton/prep float32_8\n" +
                    "  [ Fri, 23 Mar 2012 10:00:00.000000 +0000"
                    " -> Fri, 23 Mar 2012 10:00:05.000000 +0000 ]\n"
@@ -1009,6 +1016,18 @@ class TestCmdline(object):
         self.match("[ Thu, 01 Jan 2004 00:00:00.000000 +0000 -"
                    "> Sat, 01 Jan 2005 00:00:00.000000 +0000 ]\n")
 
+        # optimize
+        self.ok("insert -s 01-01-2002 -e 01-01-2004 /diff/1 /dev/null")
+        self.ok("intervals /diff/1")
+        self.match("[ Sat, 01 Jan 2000 00:00:00.000000 +0000 -"
+                   "> Thu, 01 Jan 2004 00:00:00.000000 +0000 ]\n"
+                   "[ Thu, 01 Jan 2004 00:00:00.000000 +0000 -"
+                   "> Sat, 01 Jan 2005 00:00:00.000000 +0000 ]\n")
+        self.ok("intervals /diff/1 --optimize")
+        self.ok("intervals /diff/1 -o")
+        self.match("[ Sat, 01 Jan 2000 00:00:00.000000 +0000 -"
+                   "> Sat, 01 Jan 2005 00:00:00.000000 +0000 ]\n")
+
         self.ok("destroy -R /diff/1")
         self.ok("destroy -R /diff/2")
 
@@ -1043,7 +1062,7 @@ class TestCmdline(object):
             else:
                 raise AssertionError("data not found at " + seek)
             # Verify "list" output
-            self.ok("list")
+            self.ok("list -l")
             self.match("/" + "/".join(components) + " float32_8\n")
 
         # Lots of renames

tests/test_interval.py

@@ -59,6 +59,14 @@ class TestInterval:
         self.test_interval_intersect()
         Interval = NilmdbInterval
 
+        # Other helpers in nilmdb.utils.interval
+        i = [ UtilsInterval(1,2), UtilsInterval(2,3), UtilsInterval(4,5) ]
+        eq_(list(nilmdb.utils.interval.optimize(i)),
+            [ UtilsInterval(1,3), UtilsInterval(4,5) ])
+        eq_(UtilsInterval(1234567890123456, 1234567890654321).human_string(),
+            "[ Fri, 13 Feb 2009 18:31:30.123456 -0500 -> " +
+            "Fri, 13 Feb 2009 18:31:30.654321 -0500 ]")
+
     def test_interval(self):
         # Test Interval class
         os.environ['TZ'] = "America/New_York"
@@ -226,13 +234,16 @@ class TestInterval:
             x = makeset("[--)") & 1234
 
         def do_test(a, b, c, d):
-            # a & b == c
+            # a & b == c (using nilmdb.server.interval)
             ab = IntervalSet()
             for x in b:
                 for i in (a & x):
                     ab += i
             eq_(ab,c)
 
+            # a & b == c (using nilmdb.utils.interval)
+            eq_(IntervalSet(nilmdb.utils.interval.intersection(a,b)), c)
+
             # a \ b == d
             eq_(IntervalSet(nilmdb.utils.interval.set_difference(a,b)), d)
 
@@ -302,6 +313,17 @@ class TestInterval:
         eq_(nilmdb.utils.interval.set_difference(
             a.intersection(list(c)[0]), b.intersection(list(c)[0])), d)
 
+        # Fill out test coverage for non-subsets
+        def diff2(a,b, subset):
+            return nilmdb.utils.interval._interval_math_helper(
+                a, b, (lambda a, b: b and not a), subset=subset)
+        with assert_raises(nilmdb.utils.interval.IntervalError):
+            list(diff2(a,b,True))
+        list(diff2(a,b,False))
+
+        # Empty second set
+        eq_(nilmdb.utils.interval.set_difference(a, IntervalSet()), a)
+
         # Empty second set
         eq_(nilmdb.utils.interval.set_difference(a, IntervalSet()), a)
 

tests/test_nilmdb.py

@@ -157,11 +157,14 @@ class TestServer(object):
 
     def test_server(self):
         # Make sure we can't force an exit, and test other 404 errors
-        for url in [ "/exit", "/", "/favicon.ico" ]:
+        for url in [ "/exit", "/favicon.ico" ]:
             with assert_raises(HTTPError) as e:
                 geturl(url)
             eq_(e.exception.code, 404)
 
+        # Root page
+        in_("This is NilmDB", geturl("/"))
+
         # Check version
         eq_(distutils.version.LooseVersion(getjson("/version")),
             distutils.version.LooseVersion(nilmdb.__version__))

tests/test_numpyclient.py

@@ -28,7 +28,10 @@ def setup_module():
     recursive_unlink(testdb)
 
     # Start web app on a custom port
-    test_db = nilmdb.utils.serializer_proxy(nilmdb.server.NilmDB)(testdb)
+    test_db = nilmdb.utils.serializer_proxy(nilmdb.server.NilmDB)(
+        testdb, bulkdata_args = { "file_size" : 16384,
+                                  "files_per_dir" : 3 } )
+
     test_server = nilmdb.server.Server(test_db, host = "127.0.0.1",
                                        port = 32180, stoppable = False,
                                        fast_shutdown = True,
@@ -130,6 +133,15 @@ class TestNumpyClient(object):
                                                   [4, 5]]]))
         in_("wrong number of dimensions", str(e.exception))
 
+        # Wrong number of fields
+        with assert_raises(ValueError) as e:
+            client.stream_insert_numpy("/test/1",
+                                       np.array([[0, 1, 2],
+                                                 [1, 2, 3],
+                                                 [3, 4, 5],
+                                                 [4, 5, 6]]))
+        in_("wrong number of fields", str(e.exception))
+
         # Unstructured
         client.stream_create("/test/2", "float32_8")
         client.stream_insert_numpy(
@@ -170,6 +182,17 @@ class TestNumpyClient(object):
         assert(np.array_equal(a,b))
         assert(np.array_equal(a,c))
 
+        # Make sure none of the files are greater than 16384 bytes as
+        # we configured with the bulkdata_args above.
+        datapath = os.path.join(testdb, "data")
+        for (dirpath, dirnames, filenames) in os.walk(datapath):
+            for f in filenames:
+                fn = os.path.join(dirpath, f)
+                size = os.path.getsize(fn)
+                if size > 16384:
+                    raise AssertionError(sprintf("%s is too big: %d > %d\n",
+                                                 fn, size, 16384))
+
         nilmdb.client.numpyclient.StreamInserterNumpy._max_data = old_max_data
         client.close()
 
@@ -286,8 +309,25 @@ class TestNumpyClient(object):
         with client.stream_insert_numpy_context("/empty/test", end = 950):
             pass
 
+        # Equal start and end is OK as long as there's no data
+        with assert_raises(ClientError) as e:
+            with client.stream_insert_numpy_context("/empty/test",
+                                                    start=9, end=9) as ctx:
+                ctx.insert([[9, 9]])
+                ctx.finalize()
+        in_("have data to send, but invalid start/end times", str(e.exception))
+
+        with client.stream_insert_numpy_context("/empty/test",
+                                                start=9, end=9) as ctx:
+            pass
+
+        # reusing a context object is bad
+        with assert_raises(Exception) as e:
+            ctx.insert([[9, 9]])
+
         # Try various things that might cause problems
-        with client.stream_insert_numpy_context("/empty/test", 1000, 1050):
+        with client.stream_insert_numpy_context("/empty/test",
+                                                1000, 1050) as ctx:
             ctx.finalize() # inserts [1000, 1050]
             ctx.finalize() # nothing
             ctx.finalize() # nothing

tests/test_serializer.py

@@ -62,6 +62,28 @@ class Base(object):
         eq_(self.foo.val, 20)
         eq_(self.foo.init_thread, self.foo.test_thread)
 
+class ListLike(object):
+    def __init__(self):
+        self.thread = threading.current_thread().name
+        self.foo = 0
+
+    def __iter__(self):
+        eq_(threading.current_thread().name, self.thread)
+        self.foo = 0
+        return self
+
+    def __getitem__(self, key):
+        eq_(threading.current_thread().name, self.thread)
+        return key
+
+    def next(self):
+        eq_(threading.current_thread().name, self.thread)
+        if self.foo < 5:
+            self.foo += 1
+            return self.foo
+        else:
+            raise StopIteration
+
 class TestUnserialized(Base):
     def setUp(self):
         self.foo = Foo()
@@ -84,3 +106,9 @@ class TestSerializer(Base):
         sp(sp(Foo("x"))).t()
         sp(sp(Foo)("x")).t()
         sp(sp(Foo))("x").t()
+
+    def test_iter(self):
+        sp = nilmdb.utils.serializer_proxy
+        i = sp(ListLike)()
+        eq_(list(i), [1,2,3,4,5])
+        eq_(i[3], 3)