Add nilmdb-fsck tool to check database consistency

.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

Makefile

@@ -23,6 +23,10 @@ docs:
 lint:
 	pylint --rcfile=.pylintrc nilmdb
 
+fscktest:
+#	python -c "import nilmdb.fsck; nilmdb.fsck.Fsck('/home/jim/wsgi/db').check()"
+	python -c "import nilmdb.fsck; nilmdb.fsck.Fsck('/home/jim/mnt/bucket/mnt/sharon/data/db').check()"
+
 test:
 ifeq ($(INSIDE_EMACS), t)
 # Use the slightly more flexible script

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

nilmdb/client/client.py

@@ -58,6 +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
@@ -69,6 +74,7 @@ class Client(object):
         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

nilmdb/cmdline/cmdline.py

@@ -29,6 +29,14 @@ 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))

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(
+        intervals = ( Interval(start, end) for (start, end) in
-               self.args.path, self.args.start, self.args.end, self.args.diff):
+                      self.client.stream_intervals(self.args.path,
-               printf("[ %s -> %s ]\n", time_string(start), time_string(end))
+                                                   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/fsck/.#fsck.py

@@ -0,0 +1 @@
+jim@pilot.lees.18066:1373305995

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 +1,194 @@
+# -*- coding: utf-8 -*-
+
+"""Check database consistency"""
+
+import nilmdb.utils
+import nilmdb.server
+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 progressbar
+import time
+import cPickle as pickle
+
+class FsckError(Exception):
+    def __init__(self, format, *args):
+        Exception.__init__(self, sprintf(format, *args))
+
+def log(format, *args):
+    printf(format, *args)
+
+def err(format, *args):
+    fprintf(sys.stderr, format, *args)
+
+class Progress(object):
+    def __init__(self, maxval):
+        self.bar = progressbar.ProgressBar(maxval = maxval)
+        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)
+        #now = time.time()
+        #if now - self.last_update < 0.005:
+        #   time.sleep(0.005)
+        #self.last_update = now
+
+class Fsck(object):
+
+    def __init__(self, path):
+        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")
+
+    def check(self):
+        self.check_paths()
+        self.check_sql()
+        self.check_streams()
+        log("ok\n")
+
+    def check_paths(self):
+        log("checking paths\n")
+        if not os.path.isfile(self.sqlpath):
+            raise FsckError("SQL database missing")
+        if not os.path.isdir(self.bulkpath):
+            raise FsckError("Bulk data directory missing")
+        with open(self.bulklock, "w") as lockfile:
+            if not nilmdb.utils.lock.exclusive_lock(lockfile):
+                raise FsckError('database already locked by another process')
+        self.bulk = nilmdb.server.bulkdata.BulkData(self.basepath)
+        # override must_close warning
+        if "_must_close" in dir(self.bulk):
+            del self.bulk._must_close
+
+    def check_sql(self):
+        log("checking sqlite database\n")
+
+        self.sql = sqlite3.connect(self.sqlpath)
+        with self.sql as con:
+            ver = con.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 = con.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 = con.execute("SELECT stream_id, start_time, end_time, "
+                                 "start_pos, end_pos FROM ranges")
+            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 = con.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]
+
+    def check_streams(self):
+        log("checking streams\n")
+        ids = self.stream_path.keys()
+        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\n",
+                                        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\n",
+                                            path, str(posiset), str(new))
+
+                # check bulkdata
+                self.check_bulkdata(sid, path, bulk)
+
+                continue
+                # verify we can can open it with 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))
+                    self.check_bulkdata(path, tab)
+                finally:
+                    if tab:
+                        tab.close()
+
+    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])

nilmdb/scripts/nilmdb_fsck.py

@@ -0,0 +1,23 @@
+#!/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)
+    parser.add_argument("-V", "--version", action="version",
+                        version = nilmdb.__version__)
+    parser.add_argument('-d', '--database', help = 'Database directory',
+                        default = "./db")
+    args = parser.parse_args()
+
+    nilmdb.fsck.Fsck(args.database).check()
+
+if __name__ == "__main__":
+    main()

nilmdb/server/bulkdata.py

@@ -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/server.py

@@ -27,6 +27,7 @@ from nilmdb.server.serverutil import (
     json_error_page,
     cherrypy_start,
     cherrypy_stop,
+    bool_param,
     )
 
 # Add CORS_allow tool
@@ -83,10 +84,18 @@ class Stream(NilmApp):
     # Helpers
     def _get_times(self, start_param, end_param):
         (start, end) = (None, None)
-        if start_param is not None:
+        try:
-            start = string_to_timestamp(start_param)
+            if start_param is not None:
-        if end_param is not None:
+                start = string_to_timestamp(start_param)
-            end = string_to_timestamp(end_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(
@@ -221,6 +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
@@ -345,6 +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

nilmdb/server/serverutil.py

@@ -7,6 +7,21 @@ 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."""

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):
+def _interval_math_helper(a, b, op, subset = True):
-    """
+    """Helper for set_difference, intersection functions,
-    Compute the difference (a \\ b) between the intervals in 'a' and
+    to compute interval subsets based on a math operator on ranges
-    the intervals in 'b'; i.e., the ranges that are present in 'self'
+    present in A and B.  Subsets are computed from A, or new intervals
-    but not 'other'.
+    are generated if subset = False."""
-
-    '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).
-    """
     # 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:
+            in_a = True
-                out_start = ts
         elif k == 1:
-            # start b interval
+            in_b = True
-            b_interval = i
-            if out_start is not None and out_start != ts:
-                yield a_interval.subset(out_start, ts)
-            out_start = None
         elif k == 2:
-            # end a interval
+            in_a = False
-            if out_start is not None and out_start != ts:
-                yield a_interval.subset(out_start, ts)
-            out_start = None
-            a_interval = None
         elif k == 3:
-            # end b interval
+            in_b = False
-            b_interval = None
+        include = op(in_a, in_b)
-            if a_interval:
+        if include and out_start is None:
-                out_start = ts
+            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/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

@@ -1,7 +1,14 @@
+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, unicode):
+    if not isinstance(u, text_type):
         return u
     try:
         return u.encode()
@@ -11,7 +18,7 @@ def encode(u):
 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, unicode):
+    if isinstance(s, text_type):
         return s
     try:
         return s.decode()

setup.py

@@ -126,11 +126,13 @@ 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/test_client.py

@@ -242,6 +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", "",
@@ -354,10 +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:
@@ -396,27 +405,38 @@ class TestClient(object):
                                  headers())
 
         # Extract
-        x = http.get("stream/extract",
+        x = http.get("stream/extract", { "path": "/newton/prep",
-                            { "path": "/newton/prep",
+                                         "start": "123", "end": "124" })
-                              "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",
+        x = http.get("stream/extract", { "path": "/newton/prep",
-                            { "path": "/newton/prep",
+                                         "start": "123", "end": "124",
-                              "start": "123",
+                                         "binary": "1" })
-                              "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):

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
+        passing the given input.  Save the output and exit code."""
-        exit code"""
         # printf("TZ=UTC ./nilmtool.py %s\n", arg_string)
         os.environ['NILMDB_URL'] = "http://localhost:32180/"
         class stdio_wrapper:
@@ -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("")
 
@@ -1011,6 +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")
 

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_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)