Merge branch 'binary' into HEAD

Still allow installation without it, though.

docs/wsgi.md

@@ -19,8 +19,9 @@ Then, set up Apache with a configuration like:
 
     <VirtualHost>
         WSGIScriptAlias /nilmdb /home/nilm/nilmdb.wsgi
-        WSGIProcessGroup nilmdb-server
-        WSGIDaemonProcess nilmdb-server threads=32 user=nilm group=nilm
+        WSGIApplicationGroup nilmdb-appgroup
+        WSGIProcessGroup nilmdb-procgroup
+        WSGIDaemonProcess nilmdb-procgroup threads=32 user=nilm group=nilm
 
         # Access control example:
         <Location /nilmdb>

nilmdb/client/client.py

@@ -6,6 +6,7 @@ import nilmdb.utils
 import nilmdb.client.httpclient
 from nilmdb.client.errors import ClientError
 
+import re
 import time
 import simplejson as json
 import contextlib
@@ -65,7 +66,12 @@ class Client(object):
             params["layout"] = layout
         if extended:
             params["extended"] = 1
-        return self.http.get("stream/list", params)
+        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))
 
     def stream_get_metadata(self, path, keys = None):
         params = { "path": path }
@@ -172,7 +178,7 @@ class Client(object):
         return self.http.get_gen("stream/intervals", params)
 
     def stream_extract(self, path, start = None, end = None,
-                       count = False, markup = False):
+                       count = False, markup = False, binary = False):
         """
         Extract data from a stream.  Returns a generator that yields
         lines of ASCII-formatted data that matches the database
@@ -183,6 +189,11 @@ class Client(object):
 
         Specify markup = True to include comments in the returned data
         that indicate interval starts and ends.
+
+        Specify binary = True to return chunks of raw binary data,
+        rather than lines of ASCII-formatted data.  Raw binary data
+        is always little-endian and matches the database types
+        (including a uint64 timestamp).
         """
         params = {
             "path": path,
@@ -195,7 +206,9 @@ class Client(object):
             params["count"] = 1
         if markup:
             params["markup"] = 1
-        return self.http.get_gen("stream/extract", params)
+        if binary:
+            params["binary"] = 1
+        return self.http.get_gen("stream/extract", params, binary = binary)
 
     def stream_count(self, path, start = None, end = None):
         """
@@ -313,6 +326,11 @@ class StreamInserter(object):
         part of a new interval and there may be a gap left in-between."""
         self._send_block(final = True)
 
+    def send(self):
+        """Send any data that we might have buffered up.  Does not affect
+        any other treatment of timestamps or endpoints."""
+        self._send_block(final = False)
+
     def _get_first_noncomment(self, block):
         """Return the (start, end) indices of the first full line in
         block that isn't a comment, or raise IndexError if

nilmdb/client/httpclient.py

@@ -110,7 +110,8 @@ class HTTPClient(object):
         return self._req("PUT", url, params, data)
 
     # Generator versions that return data one line at a time.
-    def _req_gen(self, method, url, query = None, body = None, headers = None):
+    def _req_gen(self, method, url, query = None, body = None,
+                 headers = None, binary = False):
         """
         Make a request and return a generator that gives back strings
         or JSON decoded lines of the body data, or raise an error if
@@ -118,16 +119,19 @@ class HTTPClient(object):
         """
         (response, isjson) = self._do_req(method, url, query, body,
                                           stream = True, headers = headers)
-        if isjson:
+        if binary:
+            for chunk in response.iter_content(chunk_size = 65536):
+                yield chunk
+        elif isjson:
             for line in response.iter_lines():
                 yield json.loads(line)
         else:
             for line in response.iter_lines():
                 yield line
 
-    def get_gen(self, url, params = None):
+    def get_gen(self, url, params = None, binary = False):
         """Simple GET (parameters in URL) returning a generator"""
-        return self._req_gen("GET", url, params)
+        return self._req_gen("GET", url, params, binary = binary)
 
     # Not much use for a POST or PUT generator, since they don't
     # return much data.

nilmdb/client/numpyclient.py

@@ -0,0 +1,77 @@
+# -*- coding: utf-8 -*-
+
+"""Provide a NumpyClient class that is based on normal Client, but has
+additional methods for extracting and inserting data via Numpy arrays."""
+
+import nilmdb.utils
+import nilmdb.client.client
+import nilmdb.client.httpclient
+from nilmdb.client.errors import ClientError
+
+import contextlib
+from nilmdb.utils.time import timestamp_to_string, string_to_timestamp
+
+import numpy
+import cStringIO
+
+def layout_to_dtype(layout):
+    ltype = layout.split('_')[0]
+    lcount = int(layout.split('_')[1])
+    if ltype.startswith('int'):
+        atype = '<i' + str(int(ltype[3:]) / 8)
+    elif ltype.startswith('uint'):
+        atype = '<u' + str(int(ltype[4:]) / 8)
+    elif ltype.startswith('float'):
+        atype = '<f' + str(int(ltype[5:]) / 8)
+    else:
+        raise ValueError("bad layout")
+    return numpy.dtype([('timestamp', '<i8'), ('data', atype, lcount)])
+
+class NumpyClient(nilmdb.client.client.Client):
+    """Subclass of nilmdb.client.Client that adds additional methods for
+    extracting and inserting data via Numpy arrays."""
+
+    def stream_extract_numpy(self, path, start = None, end = None,
+                             layout = None, maxrows = 100000,
+                             structured = False):
+        """
+        Extract data from a stream.  Returns a generator that yields
+        Numpy arrays of up to 'maxrows' of data each.
+
+        If 'layout' is None, it is read using stream_info.
+
+        If 'structured' is False, all data is converted to float64
+        and returned in a flat 2D array.  Otherwise, data is returned
+        as a structured dtype in a 1D array.
+        """
+        if layout is None:
+            streams = self.stream_list(path)
+            if len(streams) != 1:
+                raise ClientError("can't get layout for path: " + path)
+            layout = streams[0][1]
+        dtype = layout_to_dtype(layout)
+
+        def to_numpy(data):
+            a = numpy.fromstring(data, dtype)
+            if structured:
+                return a
+            return numpy.c_[a['timestamp'], a['data']]
+
+        chunks = []
+        total_len = 0
+        maxsize = dtype.itemsize * maxrows
+        for data in self.stream_extract(path, start, end, binary = True):
+            # Add this block of binary data
+            chunks.append(data)
+            total_len += len(data)
+
+            # See if we have enough to make the requested Numpy array
+            while total_len >= maxsize:
+                assembled = "".join(chunks)
+                total_len -= maxsize
+                chunks = [ assembled[maxsize:] ]
+                block = assembled[:maxsize]
+                yield to_numpy(block)
+
+        if total_len:
+            yield to_numpy("".join(chunks))

nilmdb/cmdline/cmdline.py

@@ -81,7 +81,7 @@ class Cmdline(object):
     def __init__(self, argv = None):
         self.argv = argv or sys.argv[1:]
         self.client = None
-        self.def_url = os.environ.get("NILMDB_URL", "http://localhost:12380")
+        self.def_url = os.environ.get("NILMDB_URL", "http://localhost/nilmdb/")
         self.subcmd = {}
         self.complete = Complete()
 

nilmdb/cmdline/metadata.py

@@ -9,7 +9,8 @@ def setup(self, sub):
                          a stream.
                          """,
                          usage="%(prog)s path [-g [key ...] | "
-                         "-s key=value [...] | -u key=value [...]]")
+                         "-s key=value [...] | -u key=value [...]] | "
+                         "-d [key ...]")
     cmd.set_defaults(handler = cmd_metadata)
 
     group = cmd.add_argument_group("Required arguments")
@@ -30,6 +31,9 @@ def setup(self, sub):
                      help="Update metadata using provided "
                      "key=value pairs",
                      ).completer = self.complete.meta_keyval
+    exc.add_argument("-d", "--delete", nargs="*", metavar="key",
+                     help="Delete metadata for specified keys (default all)",
+                     ).completer = self.complete.meta_key
     return cmd
 
 def cmd_metadata(self):
@@ -56,6 +60,16 @@ def cmd_metadata(self):
             handler(self.args.path, data)
         except nilmdb.client.ClientError as e:
             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
+        try:
+            data = self.client.stream_get_metadata(self.args.path, keys)
+            for key in data:
+                data[key] = ""
+            self.client.stream_update_metadata(self.args.path, data)
+        except nilmdb.client.ClientError as e:
+            self.die("error deleting metadata: %s", str(e))
     else:
         # Get (or unspecified)
         keys = self.args.get or None
@@ -64,7 +78,7 @@ def cmd_metadata(self):
         except nilmdb.client.ClientError as e:
             self.die("error getting metadata: %s", str(e))
         for key, value in sorted(data.items()):
-            # Omit nonexistant keys
+            # Print nonexistant keys as having empty value
             if value is None:
                 value = ""
             printf("%s=%s\n", key, value)

nilmdb/server/bulkdata.py

@@ -79,7 +79,12 @@ class BulkData(object):
         if Table.exists(ospath):
             raise ValueError("stream already exists at this path")
         if os.path.isdir(ospath):
-            raise ValueError("subdirs of this path already exist")
+            # Look for any files in subdirectories.  Fully empty subdirectories
+            # are OK; they might be there during a rename
+            for (root, dirs, files) in os.walk(ospath):
+                if len(files):
+                    raise ValueError(
+                        "non-empty subdirs of this path already exist")
 
     def _create_parents(self, unicodepath):
         """Verify the path name, and create parent directories if they
@@ -188,7 +193,6 @@ class BulkData(object):
         # Basic checks
         if oldospath == newospath:
             raise ValueError("old and new paths are the same")
-        self._create_check_ospath(newospath)
 
         # Move the table to a temporary location
         tmpdir = tempfile.mkdtemp(prefix = "rename-", dir = self.root)
@@ -196,6 +200,9 @@ class BulkData(object):
         os.rename(oldospath, tmppath)
 
         try:
+            # Check destination path
+            self._create_check_ospath(newospath)
+
             # Create parent dirs for new location
             self._create_parents(newunicodepath)
 
@@ -472,7 +479,7 @@ class Table(object):
             # Success, so update self.nrows accordingly
             self.nrows = tot_rows
 
-    def get_data(self, start, stop):
+    def get_data(self, start, stop, binary = False):
         """Extract data corresponding to Python range [n:m],
         and returns a formatted string"""
         if (start is None or
@@ -490,10 +497,13 @@ class Table(object):
             if count > remaining:
                 count = remaining
             f = self.file_open(subdir, filename)
-            ret.append(f.extract_string(offset, count))
+            if binary:
+                ret.append(f.extract_binary(offset, count))
+            else:
+                ret.append(f.extract_string(offset, count))
             remaining -= count
             row += count
-        return "".join(ret)
+        return b"".join(ret)
 
     def __getitem__(self, row):
         """Extract timestamps from a row, with table[n] notation."""

nilmdb/server/nilmdb.py

@@ -538,7 +538,7 @@ class NilmDB(object):
                                   dbinterval.db_endpos)
 
     def stream_extract(self, path, start = None, end = None,
-                       count = False, markup = False):
+                       count = False, markup = False, binary = False):
         """
         Returns (data, restart) tuple.
 
@@ -559,6 +559,9 @@ class NilmDB(object):
         'markup', if true, indicates that returned data should be
         marked with a comment denoting when a particular interval
         starts, and another comment when an interval ends.
+
+        'binary', if true, means to return raw binary rather than
+        ASCII-formatted data.
         """
         stream_id = self._stream_id(path)
         table = self.data.getnode(path)
@@ -569,6 +572,8 @@ class NilmDB(object):
         matched = 0
         remaining = self.max_results
         restart = None
+        if binary and (markup or count):
+            raise NilmDBError("binary mode can't be used with markup or count")
         for interval in intervals.intersection(requested):
             # Reading single rows from the table is too slow, so
             # we use two bisections to find both the starting and
@@ -593,7 +598,7 @@ class NilmDB(object):
                               timestamp_to_string(interval.start) + "\n")
 
             # Gather these results up
-            result.append(table.get_data(row_start, row_end))
+            result.append(table.get_data(row_start, row_end, binary))
 
             # Count them
             remaining -= row_end - row_start

nilmdb/server/rocket.c

@@ -527,6 +527,46 @@ err:
 	return NULL;
 }
 
+/****
+ * Extract to binary string containing raw little-endian binary data
+ */
+static PyObject *Rocket_extract_binary(Rocket *self, PyObject *args)
+{
+	long count;
+	long offset;
+
+	if (!PyArg_ParseTuple(args, "ll", &offset, &count))
+		return NULL;
+	if (!self->file) {
+		PyErr_SetString(PyExc_Exception, "no file");
+		return NULL;
+	}
+	/* Seek to target location */
+	if (fseek(self->file, offset, SEEK_SET) < 0) {
+		PyErr_SetFromErrno(PyExc_OSError);
+		return NULL;
+	}
+
+        uint8_t *str;
+        int len = count * self->binary_size;
+        str = malloc(len);
+        if (str == NULL) {
+                PyErr_SetFromErrno(PyExc_OSError);
+                return NULL;
+        }
+
+        /* Data in the file is already in the desired little-endian
+           binary format, so just read it directly. */
+        if (fread(str, self->binary_size, count, self->file) != count) {
+                free(str);
+                PyErr_SetFromErrno(PyExc_OSError);
+                return NULL;
+        }
+
+	PyObject *pystr = PyBytes_FromStringAndSize((char *)str, len);
+	free(str);
+	return pystr;
+}
 
 /****
  * Extract timestamp
@@ -600,6 +640,12 @@ static PyMethodDef Rocket_methods[] = {
 	  "Extract count rows of data from the file at offset offset.\n"
 	  "Return an ascii formatted string according to the layout" },
 
+	{ "extract_binary",
+	  (PyCFunction)Rocket_extract_binary, METH_VARARGS,
+	  "extract_binary(self, offset, count)\n\n"
+	  "Extract count rows of data from the file at offset offset.\n"
+	  "Return a raw binary string of data matching the data layout." },
+
 	{ "extract_timestamp",
 	  (PyCFunction)Rocket_extract_timestamp, METH_VARARGS,
 	  "extract_timestamp(self, offset)\n\n"

nilmdb/server/server.py

@@ -11,6 +11,7 @@ from nilmdb.utils.time import string_to_timestamp
 import cherrypy
 import sys
 import os
+import socket
 import simplejson as json
 import decorator
 import psutil
@@ -173,6 +174,21 @@ class Root(NilmApp):
 class Stream(NilmApp):
     """Stream-specific operations"""
 
+    # 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)
+        if start is not None and end is not None:
+            if start >= end:
+                raise cherrypy.HTTPError(
+                    "400 Bad Request",
+                    sprintf("start must precede end (%s >= %s)",
+                            start_param, end_param))
+        return (start, end)
+
     # /stream/list
     # /stream/list?layout=float32_8
     # /stream/list?path=/newton/prep&extended=1
@@ -301,16 +317,11 @@ class Stream(NilmApp):
         body = cherrypy.request.body.read()
 
         # Check path and get layout
-        streams = self.db.stream_list(path = path)
-        if len(streams) != 1:
-            raise cherrypy.HTTPError("404 Not Found", "No such stream")
+        if len(self.db.stream_list(path = path)) != 1:
+            raise cherrypy.HTTPError("404", "No such stream: " + path)
 
         # Check limits
-        start = string_to_timestamp(start)
-        end = string_to_timestamp(end)
-        if start >= end:
-            raise cherrypy.HTTPError("400 Bad Request",
-                                     "start must precede end")
+        (start, end) = self._get_times(start, end)
 
         # Pass the data directly to nilmdb, which will parse it and
         # raise a ValueError if there are any problems.
@@ -332,14 +343,7 @@ class Stream(NilmApp):
         the interval [start, end).  Returns the number of data points
         removed.
         """
-        if start is not None:
-            start = string_to_timestamp(start)
-        if end is not None:
-            end = string_to_timestamp(end)
-        if start is not None and end is not None:
-            if start >= end:
-                raise cherrypy.HTTPError("400 Bad Request",
-                                         "start must precede end")
+        (start, end) = self._get_times(start, end)
         total_removed = 0
         while True:
             (removed, restart) = self.db.stream_remove(path, start, end)
@@ -370,15 +374,7 @@ class Stream(NilmApp):
         Note that the response type is the non-standard
         'application/x-json-stream' for lack of a better option.
         """
-        if start is not None:
-            start = string_to_timestamp(start)
-        if end is not None:
-            end = string_to_timestamp(end)
-
-        if start is not None and end is not None:
-            if start >= end:
-                raise cherrypy.HTTPError("400 Bad Request",
-                                         "start must precede end")
+        (start, end) = self._get_times(start, end)
 
         if len(self.db.stream_list(path = path)) != 1:
             raise cherrypy.HTTPError("404", "No such stream: " + path)
@@ -404,7 +400,7 @@ class Stream(NilmApp):
     @chunked_response
     @response_type("text/plain")
     def extract(self, path, start = None, end = None,
-                count = False, markup = False):
+                count = False, markup = False, binary = False):
         """
         Extract data from backend database.  Streams the resulting
         entries as ASCII text lines separated by newlines.  This may
@@ -415,22 +411,24 @@ class Stream(NilmApp):
 
         If 'markup' is True, adds comments to the stream denoting each
         interval's start and end timestamp.
-        """
-        if start is not None:
-            start = string_to_timestamp(start)
-        if end is not None:
-            end = string_to_timestamp(end)
 
-        # Check parameters
-        if start is not None and end is not None:
-            if start >= end:
-                raise cherrypy.HTTPError("400 Bad Request",
-                                         "start must precede end")
+        If 'binary' is True, return raw binary data, rather than lines
+        of ASCII-formatted data.  Raw binary data is always
+        little-endian and matches the database types (including a
+        uint64 timestamp).
+        """
+        (start, end) = self._get_times(start, end)
 
         # Check path and get layout
-        streams = self.db.stream_list(path = path)
-        if len(streams) != 1:
-            raise cherrypy.HTTPError("404 Not Found", "No such stream")
+        if len(self.db.stream_list(path = path)) != 1:
+            raise cherrypy.HTTPError("404", "No such stream: " + path)
+
+        if binary:
+            cherrypy.response.headers['Content-Type'] = (
+                "application/octet-stream")
+            if markup or count:
+                raise cherrypy.HTTPError("400", "can't mix binary and "
+                                         "markup or count modes")
 
         @workaround_cp_bug_1200
         def content(start, end):
@@ -443,7 +441,8 @@ class Stream(NilmApp):
 
             while True:
                 (data, restart) = self.db.stream_extract(
-                    path, start, end, count = False, markup = markup)
+                    path, start, end, count = False,
+                    markup = markup, binary = binary)
                 yield data
 
                 if restart is None:
@@ -608,6 +607,11 @@ class Server(object):
     def stop(self):
         cherrypy.engine.exit()
 
+# Use a single global nilmdb.server.NilmDB and nilmdb.server.Server
+# instance since the database can only be opened once.  For this to
+# work, the web server must use only a single process and single
+# Python interpreter.  Multiple threads are OK.
+_wsgi_server = None
 def wsgi_application(dbpath, basepath): # pragma: no cover
     """Return a WSGI application object with a database at the
     specified path.
@@ -618,29 +622,33 @@ def wsgi_application(dbpath, basepath): # pragma: no cover
     is the same as the first argument to Apache's WSGIScriptAlias
     directive.
     """
-    server = [None]
     def application(environ, start_response):
-        if server[0] is None:
+        global _wsgi_server
+        if _wsgi_server is None:
             # Try to start the server
             try:
                 db = nilmdb.utils.serializer_proxy(nilmdb.server.NilmDB)(dbpath)
-                server[0] = nilmdb.server.Server(
+                _wsgi_server = nilmdb.server.Server(
                     db, embedded = True,
                     basepath = basepath.rstrip('/'))
             except Exception:
                 # Build an error message on failure
+                import pprint
                 err = sprintf("Initializing database at path '%s' failed:\n\n",
                               dbpath)
                 err += traceback.format_exc()
                 try:
                     import pwd
                     import grp
-                    err += sprintf("\nRunning as: uid=%d (%s), gid=%d (%s)\n",
+                    err += sprintf("\nRunning as: uid=%d (%s), gid=%d (%s) "
+                                   "on host %s, pid %d\n",
                                    os.getuid(), pwd.getpwuid(os.getuid())[0],
-                                   os.getgid(), grp.getgrgid(os.getgid())[0])
+                                   os.getgid(), grp.getgrgid(os.getgid())[0],
+                                   socket.gethostname(), os.getpid())
                 except ImportError:
                     pass
-        if server[0] is None:
+                err += sprintf("\nEnvironment:\n%s\n", pprint.pformat(environ))
+        if _wsgi_server is None:
             # Serve up the error with our own mini WSGI app.
             headers = [ ('Content-type', 'text/plain'),
                         ('Content-length', str(len(err))) ]
@@ -648,5 +656,5 @@ def wsgi_application(dbpath, basepath): # pragma: no cover
             return [err]
 
         # Call the normal application
-        return server[0].wsgi_application(environ, start_response)
+        return _wsgi_server.wsgi_application(environ, start_response)
     return application

nilmdb/utils/time.py

@@ -65,6 +65,14 @@ def parse_time(toparse):
     if toparse == "max":
         return max_timestamp
 
+    # If it starts with @, treat it as a NILM timestamp
+    # (integer microseconds since epoch)
+    try:
+        if toparse[0] == '@':
+            return int(toparse[1:])
+    except (ValueError, KeyError, IndexError):
+        pass
+
     # If string isn't "now" and doesn't contain at least 4 digits,
     # consider it invalid.  smartparse might otherwise accept
     # empty strings and strings with just separators.
@@ -78,14 +86,6 @@ def parse_time(toparse):
     except (ValueError, OverflowError):
         pass
 
-    # If it starts with @, treat it as a NILM timestamp
-    # (integer microseconds since epoch)
-    try:
-        if toparse[0] == '@':
-            return int(toparse[1:])
-    except (ValueError, KeyError):
-        pass
-
     # If it's parseable as a float, treat it as a Unix or NILM
     # timestamp based on its range.
     try:

setup.py

@@ -107,6 +107,7 @@ setup(name='nilmdb',
       author_email = 'jim@jtan.com',
       tests_require = [ 'nose',
                         'coverage',
+                        'numpy',
                         ],
       setup_requires = [ 'distribute',
                          ],

tests/test.order

@@ -12,6 +12,7 @@ test_interval.py
 test_bulkdata.py
 test_nilmdb.py
 test_client.py
+test_numpyclient.py
 test_cmdline.py
 
 test_*.py

tests/test_client.py

@@ -23,6 +23,7 @@ import warnings
 import resource
 import time
 import re
+import struct
 
 from testutil.helpers import *
 
@@ -293,6 +294,23 @@ class TestClient(object):
         # Test count
         eq_(client.stream_count("/newton/prep"), 14400)
 
+        # Test binary output
+        with assert_raises(ClientError) as e:
+            list(client.stream_extract("/newton/prep",
+                                       markup = True, binary = True))
+        with assert_raises(ClientError) as e:
+            list(client.stream_extract("/newton/prep",
+                                       count = True, binary = True))
+        data = "".join(client.stream_extract("/newton/prep", binary = True))
+        # Quick check using struct
+        unpacker = struct.Struct("<qffffffff")
+        out = []
+        for i in range(14400):
+            out.append(unpacker.unpack_from(data, i * unpacker.size))
+        eq_(out[0], (1332511200000000, 266568.0, 224029.0, 5161.39990234375,
+                     2525.169921875, 8350.83984375, 3724.699951171875,
+                     1355.3399658203125, 2039.0))
+
         client.close()
 
     def test_client_06_generators(self):
@@ -311,11 +329,11 @@ class TestClient(object):
 
         # Trigger a curl error in generator
         with assert_raises(ServerError) as e:
-            client.http.get_gen("http://nosuchurl/").next()
+            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/").next()
+            client.http.get_gen("http://nosuchurl.example.com./").next()
 
         # Check 404 for missing streams
         for function in [ client.stream_intervals, client.stream_extract ]:
@@ -460,6 +478,7 @@ class TestClient(object):
             ctx.update_start(109)
             ctx.insert("110 1\n")
             ctx.insert("111 1\n")
+            ctx.send()
             ctx.insert("112 1\n")
             ctx.insert("113 1\n")
             ctx.insert("114 1\n")

tests/test_cmdline.py

@@ -369,6 +369,8 @@ class TestCmdline(object):
         self.contain("No stream at path")
         self.fail("metadata /newton/nosuchstream --set foo=bar")
         self.contain("No stream at path")
+        self.fail("metadata /newton/nosuchstream --delete")
+        self.contain("No stream at path")
 
         self.ok("metadata /newton/prep")
         self.match("description=The Data\nv_scale=1.234\n")
@@ -394,6 +396,19 @@ class TestCmdline(object):
         self.fail("metadata /newton/nosuchpath")
         self.contain("No stream at path /newton/nosuchpath")
 
+        self.ok("metadata /newton/prep --delete")
+        self.ok("metadata /newton/prep --get")
+        self.match("")
+        self.ok("metadata /newton/prep --set "
+                "'description=The Data' "
+                "v_scale=1.234")
+        self.ok("metadata /newton/prep --delete v_scale")
+        self.ok("metadata /newton/prep --get")
+        self.match("description=The Data\n")
+        self.ok("metadata /newton/prep --set description=")
+        self.ok("metadata /newton/prep --get")
+        self.match("")
+
     def test_06_insert(self):
         self.ok("insert --help")
 
@@ -1038,10 +1053,12 @@ class TestCmdline(object):
         self.contain("old and new paths are the same")
         check_path("newton", "prep")
         self.fail("rename /newton/prep /newton")
-        self.contain("subdirs of this path already exist")
+        self.contain("path must contain at least one folder")
         self.fail("rename /newton/prep /newton/prep/")
         self.contain("invalid path")
-        self.ok("rename /newton/prep /newton/foo")
+        self.ok("rename /newton/prep /newton/foo/1")
+        check_path("newton", "foo", "1")
+        self.ok("rename /newton/foo/1 /newton/foo")
         check_path("newton", "foo")
         self.ok("rename /newton/foo /totally/different/thing")
         check_path("totally", "different", "thing")

tests/test_nilmdb.py

@@ -90,13 +90,16 @@ class Test00Nilmdb(object):  # named 00 so it runs first
         eq_(db.stream_get_metadata("/newton/prep"), meta1)
         eq_(db.stream_get_metadata("/newton/raw"), meta1)
 
-        # fill in some test coverage for start >= end
+        # fill in some misc. test coverage
         with assert_raises(nilmdb.server.NilmDBError):
             db.stream_remove("/newton/prep", 0, 0)
         with assert_raises(nilmdb.server.NilmDBError):
             db.stream_remove("/newton/prep", 1, 0)
         db.stream_remove("/newton/prep", 0, 1)
 
+        with assert_raises(nilmdb.server.NilmDBError):
+            db.stream_extract("/newton/prep", count = True, binary = True)
+
         db.close()
 
 class TestBlockingServer(object):

tests/test_numpyclient.py

@@ -0,0 +1,108 @@
+# -*- coding: utf-8 -*-
+
+import nilmdb.server
+import nilmdb.client
+import nilmdb.client.numpyclient
+
+from nilmdb.utils.printf import *
+from nilmdb.utils import timestamper
+from nilmdb.client import ClientError, ServerError
+from nilmdb.utils import datetime_tz
+
+from nose.plugins.skip import SkipTest
+from nose.tools import *
+from nose.tools import assert_raises
+import itertools
+import distutils.version
+
+from testutil.helpers import *
+
+import numpy as np
+
+testdb = "tests/numpyclient-testdb"
+testurl = "http://localhost:32180/"
+
+def setup_module():
+    global test_server, test_db
+    # Clear out DB
+    recursive_unlink(testdb)
+
+    # Start web app on a custom port
+    test_db = nilmdb.utils.serializer_proxy(nilmdb.server.NilmDB)(testdb)
+    test_server = nilmdb.server.Server(test_db, host = "127.0.0.1",
+                                       port = 32180, stoppable = False,
+                                       fast_shutdown = True,
+                                       force_traceback = True)
+    test_server.start(blocking = False)
+
+def teardown_module():
+    global test_server, test_db
+    # Close web app
+    test_server.stop()
+    test_db.close()
+
+class TestNumpyClient(object):
+
+    def test_numpyclient_01_basic(self):
+        # Test basic connection
+        client = nilmdb.client.numpyclient.NumpyClient(url = testurl)
+        version = client.version()
+        eq_(distutils.version.LooseVersion(version),
+            distutils.version.LooseVersion(test_server.version))
+
+        # Verify subclassing
+        assert(isinstance(client, nilmdb.client.Client))
+
+        # Layouts
+        for layout in "int8_t", "something_8", "integer_1":
+            with assert_raises(ValueError):
+                for x in client.stream_extract_numpy("/foo", layout=layout):
+                    pass
+        for layout in "int8_1", "uint8_30", "int16_20", "float64_100":
+            with assert_raises(ClientError) as e:
+                for x in client.stream_extract_numpy("/foo", layout=layout):
+                    pass
+            in_("No such stream", str(e.exception))
+
+        with assert_raises(ClientError) as e:
+            for x in client.stream_extract_numpy("/foo"):
+                pass
+        in_("can't get layout for path", str(e.exception))
+
+        client.close()
+
+    def test_numpyclient_02_extract(self):
+        client = nilmdb.client.numpyclient.NumpyClient(url = testurl)
+
+        # Insert some data as text
+        client.stream_create("/newton/prep", "float32_8")
+        testfile = "tests/data/prep-20120323T1000"
+        start = nilmdb.utils.time.parse_time("20120323T1000")
+        rate = 120
+        data = timestamper.TimestamperRate(testfile, start, rate)
+        result = client.stream_insert("/newton/prep", data,
+                                      start, start + 119999777)
+
+        # Extract Numpy arrays
+        array = None
+        pieces = 0
+        for chunk in client.stream_extract_numpy("/newton/prep", maxrows=1000):
+            pieces += 1
+            if array is not None:
+                array = np.vstack((array, chunk))
+            else:
+                array = chunk
+        eq_(array.shape, (14400, 9))
+        eq_(pieces, 15)
+
+        # Try structured
+        s = list(client.stream_extract_numpy("/newton/prep", structured = True))
+        assert(np.array_equal(np.c_[s[0]['timestamp'], s[0]['data']], array))
+
+        # Compare.  Will be close but not exact because the conversion
+        # to and from ASCII was lossy.
+        data = timestamper.TimestamperRate(testfile, start, rate)
+        actual = np.fromstring(" ".join(data), sep=' ').reshape(14400, 9)
+        assert(np.allclose(array, actual))
+
+        client.close()