Add nilmdb.utils.interval.intersection by generalizing set_difference

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/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
@@ -221,6 +222,8 @@ 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 +348,10 @@ 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):
-    """
-    Compute the difference (a \\ b) between the intervals in 'a' and
-    the intervals in 'b'; i.e., the ranges that are present in 'self'
-    but not 'other'.
-
-    'a' and 'b' must both be iterables.
-
-    Returns a generator that yields each interval in turn.
-    Output intervals are built as subsets of the intervals in the
-    first argument (a).
-    """
+def _interval_math_helper(a, b, op, subset = True):
+    """Helper for set_difference, intersection functions,
+    to compute interval subsets based on a math operator on ranges
+    present in A and B.  Subsets are computed from A, or new intervals
+    are generated if subset = False."""
     # Iterate through all starts and ends in sorted order.  Add a
     # tag to the iterator so that we can figure out which one they
     # were, after sorting.
@@ -79,28 +77,71 @@ def set_difference(a, b):
     # At each point, evaluate which type of end it is, to determine
     # how to build up the output intervals.
     a_interval = None
-    b_interval = None
+    in_a = False
+    in_b = False
     out_start = None
     for (ts, k, i) in nilmdb.utils.iterator.imerge(a_iter, b_iter):
         if k == 0:
-            # start a interval
             a_interval = i
-            if b_interval is None:
-                out_start = ts
+            in_a = True
         elif k == 1:
-            # start b interval
-            b_interval = i
-            if out_start is not None and out_start != ts:
-                yield a_interval.subset(out_start, ts)
-            out_start = None
+            in_b = True
         elif k == 2:
-            # end a interval
-            if out_start is not None and out_start != ts:
-                yield a_interval.subset(out_start, ts)
-            out_start = None
-            a_interval = None
+            in_a = False
         elif k == 3:
-            # end b interval
-            b_interval = None
-            if a_interval:
-                out_start = ts
+            in_b = False
+        include = op(in_a, in_b)
+        if include and out_start is None:
+            out_start = ts
+        elif not include:
+            if out_start is not None and out_start != ts:
+                if subset:
+                    yield a_interval.subset(out_start, ts)
+                else:
+                    yield Interval(out_start, ts)
+            out_start = None
+
+def set_difference(a, b):
+    """
+    Compute the difference (a \\ b) between the intervals in 'a' and
+    the intervals in 'b'; i.e., the ranges that are present in 'self'
+    but not 'other'.
+
+    'a' and 'b' must both be iterables.
+
+    Returns a generator that yields each interval in turn.
+    Output intervals are built as subsets of the intervals in the
+    first argument (a).
+    """
+    return _interval_math_helper(a, b, (lambda a, b: a and not b))
+
+def intersection(a, b):
+    """
+    Compute the intersection between the intervals in 'a' and the
+    intervals in 'b'; i.e., the ranges that are present in both 'a'
+    and 'b'.
+
+    'a' and 'b' must both be iterables.
+
+    Returns a generator that yields each interval in turn.
+    Output intervals are built as subsets of the intervals in the
+    first argument (a).
+    """
+    return _interval_math_helper(a, b, (lambda a, b: a and b))
+
+def optimize(it):
+    """
+    Given an iterable 'it' with intervals, optimize them by joining
+    together intervals that are adjacent in time, and return a generator
+    that yields the new intervals.
+    """
+    saved_int = None
+    for interval in it:
+        if saved_int is not None:
+            if saved_int.end == interval.start:
+                interval.start = saved_int.start
+            else:
+                yield saved_int
+        saved_int = interval
+    if saved_int is not None:
+        yield saved_int

nilmdb/utils/serializer.py

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

nilmdb/utils/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()

tests/test_client.py

@@ -354,10 +354,6 @@ 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 +392,38 @@ class TestClient(object):
                                  headers())
 
         # Extract
-        x = http.get("stream/extract",
-                            { "path": "/newton/prep",
-                              "start": "123",
-                              "end": "124" })
+        x = http.get("stream/extract", { "path": "/newton/prep",
+                                         "start": "123", "end": "124" })
         if "transfer-encoding: chunked" not in headers():
             warnings.warn("Non-chunked HTTP response for /stream/extract")
         if "content-type: text/plain;charset=utf-8" not in headers():
             raise AssertionError("/stream/extract is not text/plain:\n" +
                                  headers())
 
-        x = http.get("stream/extract",
-                            { "path": "/newton/prep",
-                              "start": "123",
-                              "end": "124",
-                              "binary": "1" })
+        x = http.get("stream/extract", { "path": "/newton/prep",
+                                         "start": "123", "end": "124",
+                                         "binary": "1" })
         if "transfer-encoding: chunked" not in headers():
             warnings.warn("Non-chunked HTTP response for /stream/extract")
         if "content-type: application/octet-stream" not in headers():
             raise AssertionError("/stream/extract is not binary:\n" +
                                  headers())
 
+        # Make sure a binary of "0" is really off
+        x = http.get("stream/extract", { "path": "/newton/prep",
+                                         "start": "123", "end": "124",
+                                         "binary": "0" })
+        if "content-type: application/octet-stream" in headers():
+                    raise AssertionError("/stream/extract is not text:\n" +
+                                         headers())
+
+        # Invalid parameters
+        with assert_raises(ClientError) as e:
+            x = http.get("stream/extract", { "path": "/newton/prep",
+                                             "start": "123", "end": "124",
+                                             "binary": "asdfasfd" })
+        in_("can't parse parameter", str(e.exception))
+
         client.close()
 
     def test_client_08_unicode(self):

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 +106,10 @@ 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)()
+        print iter(i)
+        eq_(list(i), [1,2,3,4,5])
+        eq_(i[3], 3)