Merge branch 'binary'

setup.py

@@ -61,7 +61,7 @@ setup(name='nilmtools',
       long_description = "NILM Database Tools",
       license = "Proprietary",
       author_email = 'jim@jtan.com',
-      install_requires = [ 'nilmdb >= 1.4.6',
+      install_requires = [ 'nilmdb >= 1.4.8',
                            'numpy',
                            'scipy',
                            'matplotlib',

src/copy_one.py

@@ -5,6 +5,7 @@
 
 import nilmtools.filter
 import nilmdb.client
+from nilmdb.client.numpyclient import NumpyClient
 import numpy as np
 import sys
 
@@ -27,14 +28,14 @@ def main(argv = None):
     meta = f.client_src.stream_get_metadata(f.src.path)
     f.check_dest_metadata(meta)
 
-    # Copy all rows of data as ASCII strings
-    extractor = nilmdb.client.Client(f.src.url).stream_extract
-    inserter = nilmdb.client.Client(f.dest.url).stream_insert_context
+    # Copy all rows of data using the faster Numpy interfaces
+    extractor = NumpyClient(f.src.url).stream_extract_numpy
+    inserter = NumpyClient(f.dest.url).stream_insert_numpy_context
     for i in f.intervals():
         print "Processing", f.interval_string(i)
         with inserter(f.dest.path, i.start, i.end) as insert_ctx:
-            for row in extractor(f.src.path, i.start, i.end):
-                insert_ctx.insert(row + "\n")
+            for data in extractor(f.src.path, i.start, i.end):
+                insert_ctx.insert(data)
 
 if __name__ == "__main__":
     main()

src/decimate.py

@@ -71,7 +71,7 @@ def decimate(data, interval, args, insert_function, final):
     data = data[:n,:]
 
     # Reshape it into 3D so we can process 'factor' rows at a time
-    data.shape = (n // factor, factor, m)
+    data = data.reshape(n // factor, factor, m)
 
     # Fill the result
     out = np.c_[ np.mean(data[:,:,mean_col], axis=1),

src/filter.py

@@ -4,6 +4,7 @@ from __future__ import absolute_import
 
 import nilmdb.client
 from nilmdb.client import Client
+from nilmdb.client.numpyclient import NumpyClient
 from nilmdb.utils.printf import *
 from nilmdb.utils.time import (parse_time, timestamp_to_human,
                                timestamp_to_seconds)
@@ -277,8 +278,8 @@ class Filter(object):
         """
         if args is None:
             args = []
-        extractor = Client(self.src.url).stream_extract
-        inserter = Client(self.dest.url).stream_insert_context
+        extractor = NumpyClient(self.src.url).stream_extract_numpy
+        inserter = NumpyClient(self.dest.url).stream_insert_numpy_context
 
         # Format output data.
         formatter = lambda row: " ".join([repr(x) for x in row]) + "\n"
@@ -292,22 +293,12 @@ class Filter(object):
             print "Processing", self.interval_string(interval)
             with inserter(self.dest.path,
                           interval.start, interval.end) as insert_ctx:
-                def insert_function(array):
-                    s = cStringIO.StringIO()
-                    if len(np.shape(array)) != 2:
-                        raise Exception("array must be 2-dimensional")
-                    np.savetxt(s, array)
-                    insert_ctx.insert(s.getvalue())
-
-                extract = extractor(self.src.path, interval.start, interval.end)
+                insert_function = insert_ctx.insert
                 old_array = np.array([])
-                for batched in batch(extract, rows):
-                    # Read in this batch of data.  This turns out to
-                    # be a very fast way to read and convert it (order
-                    # of magnitude faster than numpy.loadtxt)
-                    new_array = np.fromstring("\n".join(batched), sep=' ')
-                    new_array = new_array.reshape(-1, self.src.total_count)
-
+                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))