Trainola inserts into the destination stream now

This makes it easier to use the filter stuff from other code, but it's
also turning it into more of a spaghetti nightmare.  Might not be
worth continuing down this path.

Makefile

@@ -8,18 +8,21 @@ else
 	@echo "Try 'make install'"
 endif
 
-test: test_cleanup
+test: test_trainola
+
+test_trainola:
+	-nilmtool -u http://bucket/nilmdb remove -s min -e max \
+		/sharon/prep-a-matches
+	nilmtools/trainola.py "$$(cat extras/trainola-test-param.js)"
 
 test_cleanup:
 	nilmtools/cleanup.py -e extras/cleanup.cfg
 	nilmtools/cleanup.py extras/cleanup.cfg
 
 test_insert:
-	@make install >/dev/null
 	nilmtools/insert.py --file --dry-run  /test/foo </dev/null
 
 test_copy:
-	@make install >/dev/null
 	nilmtools/copy_wildcard.py -U "http://nilmdb.com/bucket/" -D /lees*
 
 /tmp/raw.dat:
@@ -29,7 +32,6 @@ test_copy:
 	--eval 'save("-ascii","/tmp/raw.dat","raw");'
 
 test_prep: /tmp/raw.dat
-	@make install >/dev/null
 	-nilmtool destroy -R /test/raw
 	-nilmtool destroy -R /test/sinefit
 	-nilmtool destroy -R /test/prep
@@ -69,4 +71,4 @@ clean::
 gitclean::
 	git clean -dXf
 
-.PHONY: all version dist sdist install clean gitclean
+.PHONY: all version dist sdist install clean gitclean test

README.txt

@@ -8,7 +8,7 @@ Prerequisites:
   sudo apt-get install python2.7 python2.7-dev python-setuptools python-pip
   sudo apt-get install python-numpy python-scipy
 
-  nilmdb (1.6.3+)
+  nilmdb (1.8.1+)
 
 Install:
 

extras/trainola-test-param.js

@@ -0,0 +1,31 @@
+{ "url": "http://bucket.mit.edu/nilmdb",
+  "dest_stream": "/sharon/prep-a-matches",
+  "stream": "/sharon/prep-a",
+  "start": 1366111383280463,
+  "end": 1366126163457797,
+  "columns": [ { "name": "P1", "index": 0 },
+               { "name": "Q1", "index": 1 },
+               { "name": "P3", "index": 2 } ],
+  "exemplars": [
+      { "name": "Boiler Pump ON",
+        "url": "http://bucket.mit.edu/nilmdb",
+        "stream": "/sharon/prep-a",
+        "start": 1366260494269078,
+        "end": 1366260608185031,
+        "dest_column": 0,
+        "columns": [ { "name": "P1", "index": 0 },
+                     { "name": "Q1", "index": 1 }
+                   ]
+      },
+      { "name": "Boiler Pump OFF",
+        "url": "http://bucket.mit.edu/nilmdb",
+        "stream": "/sharon/prep-a",
+        "start": 1366260864215764,
+        "end": 1366260870882998,
+        "dest_column": 1,
+        "columns": [ { "name": "P1", "index": 0 },
+                     { "name": "Q1", "index": 1 }
+                   ]
+      }
+  ]
+}

nilmtools/filter.py

@@ -19,6 +19,10 @@ import re
 import argparse
 import numpy as np
 import cStringIO
+import functools
+
+class ArgumentError(Exception):
+    pass
 
 class MissingDestination(Exception):
     def __init__(self, args, src, dest):
@@ -65,6 +69,70 @@ def get_stream_info(client, path):
         return None
     return StreamInfo(client.geturl(), streams[0])
 
+# Filter processing for a single interval of data.
+def process_numpy_interval(interval, extractor, inserter, warn_rows,
+                           function, args = None):
+    """For the given 'interval' of data, extract data, process it
+    through 'function', and insert the result.
+
+    'extractor' should be a function like NumpyClient.stream_extract_numpy
+    but with the the interval 'start' and 'end' as the only parameters,
+    e.g.:
+       extractor = functools.partial(NumpyClient.stream_extract_numpy,
+                                     src_path, layout = l, maxrows = m)
+
+    'inserter' should be a function like NumpyClient.stream_insert_context
+    but with the interval 'start' and 'end' as the only parameters, e.g.:
+       inserter = functools.partial(NumpyClient.stream_insert_context,
+                                    dest_path)
+
+    If 'warn_rows' is not None, print a warning to stdout when the
+    number of unprocessed rows exceeds this amount.
+
+    See process_numpy for details on 'function' and 'args'.
+    """
+    if args is None:
+        args = []
+
+    with inserter(interval.start, interval.end) as insert_ctx:
+        insert_func = insert_ctx.insert
+        old_array = np.array([])
+        for new_array in extractor(interval.start, interval.end):
+            # If we still had old data left, combine it
+            if old_array.shape[0] != 0:
+                array = np.vstack((old_array, new_array))
+            else:
+                array = new_array
+
+            # Pass the data to the user provided function
+            processed = function(array, interval, args, insert_func, False)
+
+            # Send any pending data that the user function inserted
+            insert_ctx.send()
+
+            # Save the unprocessed parts
+            if processed >= 0:
+                old_array = array[processed:]
+            else:
+                raise Exception(
+                    sprintf("%s return value %s must be >= 0",
+                            str(function), str(processed)))
+
+            # Warn if there's too much data remaining
+            if warn_rows is not None and old_array.shape[0] > warn_rows:
+                printf("warning: %d unprocessed rows in buffer\n",
+                       old_array.shape[0])
+
+        # Last call for this contiguous interval
+        if old_array.shape[0] != 0:
+            processed = function(old_array, interval, args,
+                                 insert_func, True)
+            if processed != old_array.shape[0]:
+                # Truncate the interval we're inserting at the first
+                # unprocessed data point.  This ensures that
+                # we'll not miss any data when we run again later.
+                insert_ctx.update_end(old_array[processed][0])
+
 class Filter(object):
 
     def __init__(self, parser_description = None):
@@ -134,63 +202,52 @@ class Filter(object):
         self._parser = parser
         return parser
 
-    def interval_string(self, interval):
+    def set_args(self, url, dest_url, srcpath, destpath, start, end,
-        return sprintf("[ %s -> %s ]",
+                 parsed_args = None, quiet = True):
-                       timestamp_to_human(interval.start),
+        """Set arguments directly from parameters"""
-                       timestamp_to_human(interval.end))
+        if dest_url is None:
-
+            dest_url = url
-    def parse_args(self, argv = None):
+        if url != dest_url:
-        args = self._parser.parse_args(argv)
-
-        if args.dest_url is None:
-            args.dest_url = args.url
-        if args.url != args.dest_url:
             self.interhost = True
 
-        self._client_src = Client(args.url)
+        self._client_src = Client(url)
-        self._client_dest = Client(args.dest_url)
+        self._client_dest = Client(dest_url)
 
-        if (not self.interhost) and (args.srcpath == args.destpath):
+        if (not self.interhost) and (srcpath == destpath):
-            self._parser.error("source and destination path must be different")
+            raise ArgumentError("source and destination path must be different")
 
-        # Open and print info about the streams
+        # Open the streams
-        self.src = get_stream_info(self._client_src, args.srcpath)
+        self.src = get_stream_info(self._client_src, srcpath)
         if not self.src:
-            self._parser.error("source path " + args.srcpath + " not found")
+            raise ArgumentError("source path " + srcpath + " not found")
 
-        self.dest = get_stream_info(self._client_dest, args.destpath)
+        self.dest = get_stream_info(self._client_dest, destpath)
         if not self.dest:
-            raise MissingDestination(args, self.src,
+            raise MissingDestination(parsed_args, self.src,
-                                     StreamInfo(args.dest_url, [args.destpath]))
+                                     StreamInfo(dest_url, [destpath]))
 
+        self.start = start
+        self.end = end
+
+        # Print info
+        if not quiet:
             print "Source:", self.src.string(self.interhost)
             print "  Dest:", self.dest.string(self.interhost)
 
-        if args.dry_run:
+    def parse_args(self, argv = None):
-            for interval in self.intervals():
+        """Parse arguments from a command line"""
-                print self.interval_string(interval)
+        args = self._parser.parse_args(argv)
-            raise SystemExit(0)
+
+        self.set_args(args.url, args.dest_url, args.srcpath, args.destpath,
+                      args.start, args.end, quiet = False, parsed_args = args)
 
         self.force_metadata = args.force_metadata
-
+        if args.dry_run:
-        self.start = args.start
+            for interval in self.intervals():
-        self.end = args.end
+                print interval.human_string()
-
+            raise SystemExit(0)
         return args
 
-    def _optimize_int(self, it):
-        """Join and yield adjacent intervals from the iterator 'it'"""
-        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
-
     def intervals(self):
         """Generate all the intervals that this filter should process"""
         self._using_client = True
@@ -217,12 +274,13 @@ class Filter(object):
                               self.src.path, diffpath = self.dest.path,
                               start = self.start, end = self.end) )
         # Optimize intervals: join intervals that are adjacent
-        for interval in self._optimize_int(intervals):
+        for interval in nilmdb.utils.interval.optimize(intervals):
             yield interval
         self._using_client = False
 
     # Misc helpers
-    def arg_time(self, toparse):
+    @staticmethod
+    def arg_time(toparse):
         """Parse a time string argument"""
         try:
             return nilmdb.utils.time.parse_time(toparse)
@@ -257,63 +315,6 @@ class Filter(object):
         # All good -- write the metadata in case it's not already there
         self._client_dest.stream_update_metadata(self.dest.path, data)
 
-    # Filter processing for a single interval of data.
-    def process_numpy_interval(self, interval, extractor, insert_ctx,
-                               function, args = None, rows = 100000):
-        """For the given 'interval' of data, extract data, process it
-        through 'function', and insert the result.
-
-        'extractor' should be a function like NumpyClient.stream_extract_numpy
-        'insert_ctx' should be a class like StreamInserterNumpy, with member
-        functions 'insert', 'send', and 'update_end'.
-
-        See process_numpy for details on 'function', 'args', and 'rows'.
-        """
-        if args is None:
-            args = []
-
-        insert_function = insert_ctx.insert
-        old_array = np.array([])
-        for new_array in extractor(self.src.path,
-                                   interval.start, interval.end,
-                                   layout = self.src.layout,
-                                   maxrows = rows):
-            # If we still had old data left, combine it
-            if old_array.shape[0] != 0:
-                array = np.vstack((old_array, new_array))
-            else:
-                array = new_array
-
-            # Pass it to the process function
-            processed = function(array, interval, args,
-                                 insert_function, False)
-
-            # Send any pending data
-            insert_ctx.send()
-
-            # Save the unprocessed parts
-            if processed >= 0:
-                old_array = array[processed:]
-            else:
-                raise Exception(
-                    sprintf("%s return value %s must be >= 0",
-                            str(function), str(processed)))
-
-            # Warn if there's too much data remaining
-            if old_array.shape[0] > 3 * rows:
-                printf("warning: %d unprocessed rows in buffer\n",
-                       old_array.shape[0])
-
-        # Last call for this contiguous interval
-        if old_array.shape[0] != 0:
-            processed = function(old_array, interval, args,
-                                 insert_function, True)
-            if processed != old_array.shape[0]:
-                # Truncate the interval we're inserting at the first
-                # unprocessed data point.  This ensures that
-                # we'll not miss any data when we run again later.
-                insert_ctx.update_end(old_array[processed][0])
-
     # The main filter processing method.
     def process_numpy(self, function, args = None, rows = 100000):
         """Calls process_numpy_interval for each interval that currently
@@ -352,12 +353,15 @@ class Filter(object):
         extractor = NumpyClient(self.src.url).stream_extract_numpy
         inserter = NumpyClient(self.dest.url).stream_insert_numpy_context
 
+        extractor_func = functools.partial(extractor, self.src.path,
+                                           layout = self.src.layout,
+                                           maxrows = rows)
+        inserter_func = functools.partial(inserter, self.dest.path)
+
         for interval in self.intervals():
-            print "Processing", self.interval_string(interval)
+            print "Processing", interval.human_string()
-            with inserter(self.dest.path,
+            process_numpy_interval(interval, extractor_func, inserter_func,
-                          interval.start, interval.end) as insert_ctx:
+                                   rows * 3, function, args)
-                self.process_numpy_interval(interval, extractor, insert_ctx,
-                                            function, args, rows)
 
 def main(argv = None):
     # This is just a dummy function; actual filters can use the other
@@ -366,7 +370,7 @@ def main(argv = None):
     parser = f.setup_parser()
     args = f.parse_args(argv)
     for i in f.intervals():
-        print "Generic filter: need to handle", f.interval_string(i)
+        print "Generic filter: need to handle", i.human_string()
 
 if __name__ == "__main__":
     main()

nilmtools/trainola.py

@@ -0,0 +1,279 @@
+#!/usr/bin/python
+
+from nilmdb.utils.printf import *
+import nilmdb.client
+import nilmtools.filter
+from nilmdb.utils.time import (timestamp_to_human,
+                               timestamp_to_seconds,
+                               seconds_to_timestamp)
+from nilmdb.utils.interval import Interval
+
+import numpy as np
+import scipy
+import scipy.signal
+from numpy.core.umath_tests import inner1d
+import nilmrun
+from collections import OrderedDict
+import sys
+import functools
+import collections
+
+class DataError(ValueError):
+    pass
+
+def build_column_mapping(colinfo, streaminfo):
+    """Given the 'columns' list from the JSON data, verify and
+    pull out a dictionary mapping for the column names/numbers."""
+    columns = OrderedDict()
+    for c in colinfo:
+        if (c['name'] in columns.keys() or
+            c['index'] in columns.values()):
+            raise DataError("duplicated columns")
+        if (c['index'] < 0 or c['index'] >= streaminfo.layout_count):
+            raise DataError("bad column number")
+        columns[c['name']] = c['index']
+    if not len(columns):
+        raise DataError("no columns")
+    return columns
+
+class Exemplar(object):
+    def __init__(self, exinfo, min_rows = 10, max_rows = 100000):
+        """Given a dictionary entry from the 'exemplars' input JSON,
+        verify the stream, columns, etc.  Then, fetch all the data
+        into self.data."""
+
+        self.name = exinfo['name']
+        self.url = exinfo['url']
+        self.stream = exinfo['stream']
+        self.start = exinfo['start']
+        self.end = exinfo['end']
+        self.dest_column = exinfo['dest_column']
+
+        # Get stream info
+        self.client = nilmdb.client.numpyclient.NumpyClient(self.url)
+        self.info = nilmtools.filter.get_stream_info(self.client, self.stream)
+
+        # Build up name => index mapping for the columns
+        self.columns = build_column_mapping(exinfo['columns'], self.info)
+
+        # Count points
+        self.count = self.client.stream_count(self.stream, self.start, self.end)
+
+        # Verify count
+        if self.count == 0:
+            raise DataError("No data in this exemplar!")
+        if self.count < min_rows:
+            raise DataError("Too few data points: " + str(self.count))
+        if self.count > max_rows:
+            raise DataError("Too many data points: " + str(self.count))
+
+        # Extract the data
+        datagen = self.client.stream_extract_numpy(self.stream,
+                                                   self.start, self.end,
+                                                   self.info.layout,
+                                                   maxrows = self.count)
+        self.data = list(datagen)[0]
+
+        # Discard timestamp
+        self.data = self.data[:,1:]
+
+        # Subtract the mean from each column
+        self.data = self.data - self.data.mean(axis=0)
+
+        # Get scale factors for each column by computing dot product
+        # of each column with itself.
+        self.scale = inner1d(self.data.T, self.data.T)
+
+        # Ensure a minimum (nonzero) scale and convert to list
+        self.scale = np.maximum(self.scale, [1e-9]).tolist()
+
+    def __str__(self):
+        return sprintf("\"%s\" %s [%s] %s rows",
+                       self.name, self.stream, ",".join(self.columns.keys()),
+                       self.count)
+
+def peak_detect(data, delta):
+    """Simple min/max peak detection algorithm, taken from my code
+    in the disagg.m from the 10-8-5 paper"""
+    mins = [];
+    maxs = [];
+    cur_min = (None, np.inf)
+    cur_max = (None, -np.inf)
+    lookformax = False
+    for (n, p) in enumerate(data):
+        if p > cur_max[1]:
+            cur_max = (n, p)
+        if p < cur_min[1]:
+            cur_min = (n, p)
+        if lookformax:
+            if p < (cur_max[1] - delta):
+                maxs.append(cur_max)
+                cur_min = (n, p)
+                lookformax = False
+        else:
+            if p > (cur_min[1] + delta):
+                mins.append(cur_min)
+                cur_max = (n, p)
+                lookformax = True
+    return (mins, maxs)
+
+def trainola_matcher(data, interval, args, insert_func, final_chunk):
+    """Perform cross-correlation match"""
+    ( src_columns, dest_count, exemplars ) = args
+    nrows = data.shape[0]
+
+    # We want at least 10% more points than the widest exemplar.
+    widest = max([ x.count for x in exemplars ])
+    if (widest * 1.1) > nrows:
+        return 0
+
+    # This is how many points we'll consider valid in the
+    # cross-correlation.
+    valid = nrows + 1 - widest
+    matches = collections.defaultdict(list)
+
+    # Try matching against each of the exemplars
+    for e in exemplars:
+        corrs = []
+
+        # Compute cross-correlation for each column
+        for col_name in e.columns:
+            a = data[:, src_columns[col_name] + 1]
+            b = e.data[:, e.columns[col_name]]
+            corr = scipy.signal.fftconvolve(a, np.flipud(b), 'valid')[0:valid]
+
+            # Scale by the norm of the exemplar
+            corr = corr / e.scale[e.columns[col_name]]
+            corrs.append(corr)
+
+        # Find the peaks using the column with the largest amplitude
+        biggest = e.scale.index(max(e.scale))
+        peaks_minmax = peak_detect(corrs[biggest], 0.1)
+        peaks = [ p[0] for p in peaks_minmax[1] ]
+
+        # Now look at every peak
+        for row in peaks:
+            # Correlation for each column must be close enough to 1.
+            for (corr, scale) in zip(corrs, e.scale):
+                # The accepted distance from 1 is based on the relative
+                # amplitude of the column.  Use a linear mapping:
+                # scale 1.0 -> distance 0.1
+                # scale 0.0 -> distance 1.0
+                distance = 1 - 0.9 * (scale / e.scale[biggest])
+                if abs(corr[row] - 1) > distance:
+                    # No match
+                    break
+            else:
+                # Successful match
+                matches[row].append(e)
+
+    # Insert matches into destination stream.
+    matched_rows = sorted(matches.keys())
+    out = np.zeros((len(matched_rows), dest_count + 1))
+
+    for n, row in enumerate(matched_rows):
+        # Fill timestamp
+        out[n][0] = data[row, 0]
+
+        # Mark matched exemplars
+        for exemplar in matches[row]:
+            out[n, exemplar.dest_column + 1] = 1.0
+
+    # Insert it
+    insert_func(out)
+
+    # Return how many rows we processed
+    return max(valid, 0)
+
+def trainola(conf):
+    print "Trainola", nilmtools.__version__
+
+    # Load main stream data
+    url = conf['url']
+    src_path = conf['stream']
+    dest_path = conf['dest_stream']
+    start = conf['start']
+    end = conf['end']
+
+    # Get info for the src and dest streams
+    src_client = nilmdb.client.numpyclient.NumpyClient(url)
+    src = nilmtools.filter.get_stream_info(src_client, src_path)
+    if not src:
+        raise DataError("source path '" + src_path + "' does not exist")
+    src_columns = build_column_mapping(conf['columns'], src)
+
+    dest_client = nilmdb.client.numpyclient.NumpyClient(url)
+    dest = nilmtools.filter.get_stream_info(dest_client, dest_path)
+    if not dest:
+        raise DataError("destination path '" + dest_path + "' does not exist")
+
+    printf("Source:\n")
+    printf("  %s [%s]\n", src.path, ",".join(src_columns.keys()))
+    printf("Destination:\n")
+    printf("  %s (%s columns)\n", dest.path, dest.layout_count)
+
+    # Pull in the exemplar data
+    exemplars = []
+    for n, exinfo in enumerate(conf['exemplars']):
+        printf("Loading exemplar %d:\n", n)
+        e = Exemplar(exinfo)
+        col = e.dest_column
+        if col < 0 or col >= dest.layout_count:
+            raise DataError(sprintf("bad destination column number %d\n" +
+                                    "dest stream only has 0 through %d",
+                                    col, dest.layout_count - 1))
+        printf("  %s, output column %d\n", str(e), col)
+        exemplars.append(e)
+    if len(exemplars) == 0:
+        raise DataError("missing exemplars")
+
+    # Verify that the exemplar columns are all represented in the main data
+    for n, ex in enumerate(exemplars):
+        for col in ex.columns:
+            if col not in src_columns:
+                raise DataError(sprintf("Exemplar %d column %s is not "
+                                        "available in source data", n, col))
+
+    # Figure out which intervals we should process
+    intervals = ( Interval(s, e) for (s, e) in
+                  src_client.stream_intervals(src_path,
+                                              diffpath = dest_path,
+                                              start = start, end = end) )
+    intervals = nilmdb.utils.interval.optimize(intervals)
+
+    # Do the processing
+    rows = 100000
+    extractor = functools.partial(src_client.stream_extract_numpy,
+                                  src.path, layout = src.layout, maxrows = rows)
+    inserter = functools.partial(dest_client.stream_insert_numpy_context,
+                                 dest.path)
+    for interval in intervals:
+        printf("Processing interval:\n")
+        printf("  %s\n", interval.human_string())
+        nilmtools.filter.process_numpy_interval(
+            interval, extractor, inserter, rows * 3,
+            trainola_matcher, (src_columns, dest.layout_count, exemplars))
+
+    return "done"
+
+def main(argv = None):
+    import simplejson as json
+    import sys
+
+    if argv is None:
+        argv = sys.argv[1:]
+    if len(argv) != 1:
+        raise DataError("need one argument, either a dictionary or JSON string")
+
+    try:
+        # Passed in a JSON string (e.g. on the command line)
+        conf = json.loads(argv[0])
+    except TypeError as e:
+        # Passed in the config dictionary (e.g. from NilmRun)
+        conf = argv[0]
+
+    return trainola(conf)
+
+if __name__ == "__main__":
+    main()
+

setup.py

@@ -61,7 +61,7 @@ setup(name='nilmtools',
       long_description = "NILM Database Tools",
       license = "Proprietary",
       author_email = 'jim@jtan.com',
-      install_requires = [ 'nilmdb >= 1.6.3',
+      install_requires = [ 'nilmdb >= 1.8.1',
                            'numpy',
                            'scipy',
                            #'matplotlib',
@@ -79,6 +79,7 @@ setup(name='nilmtools',
               'nilm-sinefit = nilmtools.sinefit:main',
               'nilm-cleanup = nilmtools.cleanup:main',
               'nilm-median = nilmtools.median:main',
+              'nilm-trainola = nilmtools.trainola:main',
               ],
           },
       zip_safe = False,