Improve boolean HTTP parameter handling

This will help with implementing nilmrun.

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/httpclient.py

@@ -123,14 +123,36 @@ class HTTPClient(object):
         """
         (response, isjson) = self._do_req(method, url, query, body,
                                           stream = True, headers = headers)
+
+        # Like the iter_lines function in Requests, but only splits on
+        # the specified line ending.
+        def lines(source, ending):
+            pending = None
+            for chunk in source:
+                if pending is not None:
+                    chunk = pending + chunk
+                tmp = chunk.split(ending)
+                lines = tmp[:-1]
+                if chunk.endswith(ending):
+                    pending = None
+                else:
+                    pending = tmp[-1]
+                for line in lines:
+                    yield line
+            if pending is not None: # pragma: no cover (missing newline)
+                yield pending
+
+        # Yield the chunks or lines as requested
         if binary:
             for chunk in response.iter_content(chunk_size = 65536):
                 yield chunk
         elif isjson:
-            for line in response.iter_lines():
+            for line in lines(response.iter_content(chunk_size = 1),
+                              ending = '\r\n'):
                 yield json.loads(line)
         else:
-            for line in response.iter_lines():
+            for line in lines(response.iter_content(chunk_size = 65536),
+                              ending = '\n'):
                 yield line
 
     def get_gen(self, url, params = None, binary = False):

nilmdb/cmdline/cmdline.py

@@ -72,10 +72,16 @@ class Complete(object): # pragma: no cover
         path = parsed_args.path
         if not path:
             return []
-        return ( self.escape(k + '=' + v)
+        results = []
-                 for (k,v) in client.stream_get_metadata(path).iteritems()
+        # prefix comes in as UTF-8, but results need to be Unicode,
-                 if k.startswith(prefix) )
+        # weird.  Still doesn't work in all cases, but that's bugs in
-
+        # argcomplete.
+        prefix = nilmdb.utils.unicode.decode(prefix)
+        for (k,v) in client.stream_get_metadata(path).iteritems():
+            kv = self.escape(k + '=' + v)
+            if kv.startswith(prefix):
+                results.append(kv)
+        return results
 
 class Cmdline(object):
 

nilmdb/cmdline/info.py

@@ -21,5 +21,8 @@ def cmd_info(self):
     printf("Server URL: %s\n", self.client.geturl())
     dbinfo = self.client.dbinfo()
     printf("Server database path: %s\n", dbinfo["path"])
-    printf("Server database size: %s\n", human_size(dbinfo["size"]))
+    for (desc, field) in [("used by NilmDB", "size"),
-    printf("Server database free space: %s\n", human_size(dbinfo["free"]))
+                          ("used by other", "other"),
+                          ("reserved", "reserved"),
+                          ("free", "free")]:
+        printf("Server disk space %s: %s\n", desc, human_size(dbinfo[field]))

nilmdb/cmdline/metadata.py

@@ -41,10 +41,10 @@ def cmd_metadata(self):
     if self.args.set is not None or self.args.update is not None:
         # Either set, or update
         if self.args.set is not None:
-            keyvals = self.args.set
+            keyvals = map(nilmdb.utils.unicode.decode, self.args.set)
             handler = self.client.stream_set_metadata
         else:
-            keyvals = self.args.update
+            keyvals = map(nilmdb.utils.unicode.decode, self.args.update)
             handler = self.client.stream_update_metadata
 
         # Extract key=value pairs
@@ -62,7 +62,9 @@ def cmd_metadata(self):
             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
+        keys = None
+        if self.args.delete:
+            keys = map(nilmdb.utils.unicode.decode, self.args.delete)
         try:
             data = self.client.stream_get_metadata(self.args.path, keys)
             for key in data:
@@ -72,7 +74,9 @@ def cmd_metadata(self):
             self.die("error deleting metadata: %s", str(e))
     else:
         # Get (or unspecified)
-        keys = self.args.get or None
+        keys = None
+        if self.args.get:
+            keys = map(nilmdb.utils.unicode.decode, self.args.get)
         try:
             data = self.client.stream_get_metadata(self.args.path, keys)
         except nilmdb.client.ClientError as e:
@@ -81,4 +85,6 @@ def cmd_metadata(self):
             # Print nonexistant keys as having empty value
             if value is None:
                 value = ""
-            printf("%s=%s\n", key, value)
+            printf("%s=%s\n",
+                   nilmdb.utils.unicode.encode(key),
+                   nilmdb.utils.unicode.encode(value))

nilmdb/server/bulkdata.py

@@ -19,8 +19,8 @@ from . import rocket
 
 # Up to 256 open file descriptors at any given time.
 # These variables are global so they can be used in the decorator arguments.
-table_cache_size = 16
+table_cache_size = 32
-fd_cache_size = 16
+fd_cache_size = 8
 
 @nilmdb.utils.must_close(wrap_verify = False)
 class BulkData(object):

nilmdb/server/nilmdb.py

@@ -176,7 +176,7 @@ class NilmDB(object):
             raise NilmDBError("start must precede end")
         return (start, end)
 
-    @nilmdb.utils.lru_cache(size = 16)
+    @nilmdb.utils.lru_cache(size = 64)
     def _get_intervals(self, stream_id):
         """
         Return a mutable IntervalSet corresponding to the given stream ID.

nilmdb/server/rocket.c

@@ -5,6 +5,9 @@
 #include <ctype.h>
 #include <stdint.h>
 
+#define __STDC_FORMAT_MACROS
+#include <inttypes.h>
+
 /* Values missing from stdint.h */
 #define UINT8_MIN 0
 #define UINT16_MIN 0
@@ -19,13 +22,6 @@
 
 typedef int64_t timestamp_t;
 
-/* This code probably needs to be double-checked for the case where
-   sizeof(long) != 8, so enforce that here with something that will
-   fail at build time.  We assume that the python integer type can
-   hold an int64_t. */
-const static char __long_ok[1 - 2*!(sizeof(int64_t) ==
-				    sizeof(long int))] = { 0 };
-
 /* Somewhat arbitrary, just so we can use fixed sizes for strings
    etc. */
 static const int MAX_LAYOUT_COUNT = 1024;
@@ -58,7 +54,7 @@ static PyObject *raise_str(int line, int col, int code, const char *string)
 static PyObject *raise_int(int line, int col, int code, int64_t num)
 {
 	PyObject *o;
-	o = Py_BuildValue("(iiil)", line, col, code, num);
+	o = Py_BuildValue("(iiiL)", line, col, code, (long long)num);
 	if (o != NULL) {
 		PyErr_SetObject(ParseError, o);
 		Py_DECREF(o);
@@ -249,11 +245,11 @@ static PyObject *Rocket_get_file_size(Rocket *self)
 /****
  * Append from string
  */
-static inline long int strtol10(const char *nptr, char **endptr) {
+static inline long int strtoll10(const char *nptr, char **endptr) {
-	return strtol(nptr, endptr, 10);
+	return strtoll(nptr, endptr, 10);
 }
-static inline long int strtoul10(const char *nptr, char **endptr) {
+static inline long int strtoull10(const char *nptr, char **endptr) {
-	return strtoul(nptr, endptr, 10);
+	return strtoull(nptr, endptr, 10);
 }
 
 /* .append_string(count, data, offset, linenum, start, end, last_timestamp) */
@@ -264,6 +260,7 @@ static PyObject *Rocket_append_string(Rocket *self, PyObject *args)
 	int offset;
 	const char *linestart;
 	int linenum;
+        long long ll1, ll2, ll3;
 	timestamp_t start;
 	timestamp_t end;
 	timestamp_t last_timestamp;
@@ -280,10 +277,13 @@ static PyObject *Rocket_append_string(Rocket *self, PyObject *args)
 	   but we need the null termination for strto*.  If we had
 	   strnto* that took a length, we could use t# and not require
 	   a copy. */
-	if (!PyArg_ParseTuple(args, "isiilll:append_string", &count,
+	if (!PyArg_ParseTuple(args, "isiiLLL:append_string", &count,
 			      &data, &offset, &linenum,
-			      &start, &end, &last_timestamp))
+			      &ll1, &ll2, &ll3))
 		return NULL;
+        start = ll1;
+        end = ll2;
+        last_timestamp = ll3;
 
 	/* Skip spaces, but don't skip over a newline. */
 #define SKIP_BLANK(buf) do {			\
@@ -372,14 +372,14 @@ static PyObject *Rocket_append_string(Rocket *self, PyObject *args)
 				goto extra_data_on_line;		\
 			break
 
-			CS(INT8,   strtol10,  t64.i, t8.i,  t8.u,         , 1);
+			CS(INT8,   strtoll10,  t64.i, t8.i,  t8.u,         , 1);
-			CS(UINT8,  strtoul10, t64.u, t8.u,  t8.u,         , 1);
+			CS(UINT8,  strtoull10, t64.u, t8.u,  t8.u,         , 1);
-			CS(INT16,  strtol10,  t64.i, t16.i, t16.u, le16toh, 2);
+			CS(INT16,  strtoll10,  t64.i, t16.i, t16.u, le16toh, 2);
-			CS(UINT16, strtoul10, t64.u, t16.u, t16.u, le16toh, 2);
+			CS(UINT16, strtoull10, t64.u, t16.u, t16.u, le16toh, 2);
-			CS(INT32,  strtol10,  t64.i, t32.i, t32.u, le32toh, 4);
+			CS(INT32,  strtoll10,  t64.i, t32.i, t32.u, le32toh, 4);
-			CS(UINT32, strtoul10, t64.u, t32.u, t32.u, le32toh, 4);
+			CS(UINT32, strtoull10, t64.u, t32.u, t32.u, le32toh, 4);
-			CS(INT64,  strtol10,  t64.i, t64.i, t64.u, le64toh, 8);
+			CS(INT64,  strtoll10,  t64.i, t64.i, t64.u, le64toh, 8);
-			CS(UINT64, strtoul10, t64.u, t64.u, t64.u, le64toh, 8);
+			CS(UINT64, strtoull10, t64.u, t64.u, t64.u, le64toh, 8);
 			CS(FLOAT32, strtod,   t64.d, t32.f, t32.u, le32toh, 4);
 			CS(FLOAT64, strtod,   t64.d, t64.d, t64.u, le64toh, 8);
 #undef CS
@@ -397,7 +397,8 @@ static PyObject *Rocket_append_string(Rocket *self, PyObject *args)
 	/* Build return value and return */
 	offset = buf - data;
 	PyObject *o;
-	o = Py_BuildValue("(iili)", written, offset, last_timestamp, linenum);
+	o = Py_BuildValue("(iiLi)", written, offset,
+                          (long long)last_timestamp, linenum);
 	return o;
 err:
 	PyErr_SetFromErrno(PyExc_OSError);
@@ -431,14 +432,18 @@ static PyObject *Rocket_append_binary(Rocket *self, PyObject *args)
         int data_len;
         int linenum;
 	int offset;
+        long long ll1, ll2, ll3;
 	timestamp_t start;
 	timestamp_t end;
 	timestamp_t last_timestamp;
 
-	if (!PyArg_ParseTuple(args, "it#iilll:append_binary",
+	if (!PyArg_ParseTuple(args, "it#iiLLL:append_binary",
                               &count, &data, &data_len, &offset,
-                              &linenum, &start, &end, &last_timestamp))
+                              &linenum, &ll1, &ll2, &ll3))
 		return NULL;
+        start = ll1;
+        end = ll2;
+        last_timestamp = ll3;
 
         /* Advance to offset */
         if (offset > data_len)
@@ -476,8 +481,8 @@ static PyObject *Rocket_append_binary(Rocket *self, PyObject *args)
 
 	/* Build return value and return */
 	PyObject *o;
-	o = Py_BuildValue("(iili)", rows, offset + rows * self->binary_size,
+	o = Py_BuildValue("(iiLi)", rows, offset + rows * self->binary_size,
-                          last_timestamp, linenum);
+                          (long long)last_timestamp, linenum);
 	return o;
 }
 
@@ -534,7 +539,7 @@ static PyObject *Rocket_extract_string(Rocket *self, PyObject *args)
 		if (fread(&t64.u, 8, 1, self->file) != 1)
 			goto err;
 		t64.u = le64toh(t64.u);
-		ret = sprintf(&str[len], "%ld", t64.i);
+		ret = sprintf(&str[len], "%" PRId64, t64.i);
 		if (ret <= 0)
 			goto err;
 		len += ret;
@@ -556,14 +561,14 @@ static PyObject *Rocket_extract_string(Rocket *self, PyObject *args)
 				len += ret;				\
 			}						\
 			break
-			CASE(INT8,   "%hhd",   t8.i,  t8.u,         , 1);
+			CASE(INT8,   "%" PRId8,  t8.i,  t8.u,         , 1);
-			CASE(UINT8,  "%hhu",   t8.u,  t8.u,         , 1);
+			CASE(UINT8,  "%" PRIu8,  t8.u,  t8.u,         , 1);
-			CASE(INT16,  "%hd",    t16.i, t16.u, le16toh, 2);
+			CASE(INT16,  "%" PRId16, t16.i, t16.u, le16toh, 2);
-			CASE(UINT16, "%hu",    t16.u, t16.u, le16toh, 2);
+			CASE(UINT16, "%" PRIu16, t16.u, t16.u, le16toh, 2);
-			CASE(INT32,  "%d",     t32.i, t32.u, le32toh, 4);
+			CASE(INT32,  "%" PRId32, t32.i, t32.u, le32toh, 4);
-			CASE(UINT32, "%u",     t32.u, t32.u, le32toh, 4);
+			CASE(UINT32, "%" PRIu32, t32.u, t32.u, le32toh, 4);
-			CASE(INT64,  "%ld",    t64.i, t64.u, le64toh, 8);
+			CASE(INT64,  "%" PRId64, t64.i, t64.u, le64toh, 8);
-			CASE(UINT64, "%lu",    t64.u, t64.u, le64toh, 8);
+			CASE(UINT64, "%" PRIu64, t64.u, t64.u, le64toh, 8);
 			/* These next two are a bit debatable.  floats
 			   are 6-9 significant figures, so we print 7.
 			   Doubles are 15-19, so we print 17.  This is
@@ -653,7 +658,7 @@ static PyObject *Rocket_extract_timestamp(Rocket *self, PyObject *args)
 
 	/* Convert and return */
 	t64.u = le64toh(t64.u);
-	return Py_BuildValue("l", t64.i);
+	return Py_BuildValue("L", (long long)t64.i);
 }
 
 /****

nilmdb/server/server.py

@@ -17,126 +17,26 @@ import decorator
 import psutil
 import traceback
 
+from nilmdb.server.serverutil import (
+    chunked_response,
+    response_type,
+    workaround_cp_bug_1200,
+    exception_to_httperror,
+    CORS_allow,
+    json_to_request_params,
+    json_error_page,
+    cherrypy_start,
+    cherrypy_stop,
+    bool_param,
+    )
+
+# Add CORS_allow tool
+cherrypy.tools.CORS_allow = cherrypy.Tool('on_start_resource', CORS_allow)
+
 class NilmApp(object):
     def __init__(self, db):
         self.db = db
 
-# Decorators
-def chunked_response(func):
-    """Decorator to enable chunked responses."""
-    # Set this to False to get better tracebacks from some requests
-    # (/stream/extract, /stream/intervals).
-    func._cp_config = { 'response.stream': True }
-    return func
-
-def response_type(content_type):
-    """Return a decorator-generating function that sets the
-    response type to the specified string."""
-    def wrapper(func, *args, **kwargs):
-        cherrypy.response.headers['Content-Type'] = content_type
-        return func(*args, **kwargs)
-    return decorator.decorator(wrapper)
-
-@decorator.decorator
-def workaround_cp_bug_1200(func, *args, **kwargs): # pragma: no cover
-    """Decorator to work around CherryPy bug #1200 in a response
-    generator.
-
-    Even if chunked responses are disabled, LookupError or
-    UnicodeError exceptions may still be swallowed by CherryPy due to
-    bug #1200.  This throws them as generic Exceptions instead so that
-    they make it through.
-    """
-    exc_info = None
-    try:
-        for val in func(*args, **kwargs):
-            yield val
-    except (LookupError, UnicodeError):
-        # Re-raise it, but maintain the original traceback
-        exc_info = sys.exc_info()
-        new_exc = Exception(exc_info[0].__name__ + ": " + str(exc_info[1]))
-        raise new_exc, None, exc_info[2]
-    finally:
-        del exc_info
-
-def exception_to_httperror(*expected):
-    """Return a decorator-generating function that catches expected
-    errors and throws a HTTPError describing it instead.
-
-        @exception_to_httperror(NilmDBError, ValueError)
-        def foo():
-            pass
-    """
-    def wrapper(func, *args, **kwargs):
-        exc_info = None
-        try:
-            return func(*args, **kwargs)
-        except expected:
-            # Re-raise it, but maintain the original traceback
-            exc_info = sys.exc_info()
-            new_exc = cherrypy.HTTPError("400 Bad Request", str(exc_info[1]))
-            raise new_exc, None, exc_info[2]
-        finally:
-            del exc_info
-    # We need to preserve the function's argspecs for CherryPy to
-    # handle argument errors correctly.  Decorator.decorator takes
-    # care of that.
-    return decorator.decorator(wrapper)
-
-# Custom CherryPy tools
-
-def CORS_allow(methods):
-    """This does several things:
-
-    Handles CORS preflight requests.
-    Adds Allow: header to all requests.
-    Raise 405 if request.method not in method.
-
-    It is similar to cherrypy.tools.allow, with the CORS stuff added.
-    """
-    request = cherrypy.request.headers
-    response = cherrypy.response.headers
-
-    if not isinstance(methods, (tuple, list)): # pragma: no cover
-        methods = [ methods ]
-    methods = [ m.upper() for m in methods if m ]
-    if not methods: # pragma: no cover
-        methods = [ 'GET', 'HEAD' ]
-    elif 'GET' in methods and 'HEAD' not in methods: # pragma: no cover
-        methods.append('HEAD')
-    response['Allow'] = ', '.join(methods)
-
-    # Allow all origins
-    if 'Origin' in request:
-        response['Access-Control-Allow-Origin'] = request['Origin']
-
-    # If it's a CORS request, send response.
-    request_method = request.get("Access-Control-Request-Method", None)
-    request_headers = request.get("Access-Control-Request-Headers", None)
-    if (cherrypy.request.method == "OPTIONS" and
-        request_method and request_headers):
-        response['Access-Control-Allow-Headers'] = request_headers
-        response['Access-Control-Allow-Methods'] = ', '.join(methods)
-        # Try to stop further processing and return a 200 OK
-        cherrypy.response.status = "200 OK"
-        cherrypy.response.body = ""
-        cherrypy.request.handler = lambda: ""
-        return
-
-    # Reject methods that were not explicitly allowed
-    if cherrypy.request.method not in methods:
-        raise cherrypy.HTTPError(405)
-
-cherrypy.tools.CORS_allow = cherrypy.Tool('on_start_resource', CORS_allow)
-
-# Helper for json_in tool to process JSON data into normal request
-# parameters.
-def json_to_request_params(body):
-    cherrypy.lib.jsontools.json_processor(body)
-    if not isinstance(cherrypy.request.json, dict):
-        raise cherrypy.HTTPError(415)
-    cherrypy.request.params.update(cherrypy.request.json)
-
 # CherryPy apps
 class Root(NilmApp):
     """Root application for NILM database"""
@@ -147,7 +47,10 @@ class Root(NilmApp):
     # /
     @cherrypy.expose
     def index(self):
-        raise cherrypy.NotFound()
+        cherrypy.response.headers['Content-Type'] = 'text/plain'
+        msg = sprintf("This is NilmDB version %s, running on host %s.\n",
+                      nilmdb.__version__, socket.getfqdn())
+        return msg
 
     # /favicon.ico
     @cherrypy.expose
@@ -167,9 +70,13 @@ class Root(NilmApp):
         """Return a dictionary with the database path,
         size of the database in bytes, and free disk space in bytes"""
         path = self.db.get_basepath()
+        usage = psutil.disk_usage(path)
+        dbsize = nilmdb.utils.du(path)
         return { "path": path,
-                 "size": nilmdb.utils.du(path),
+                 "size": dbsize,
-                 "free": psutil.disk_usage(path).free }
+                 "other": usage.used - dbsize,
+                 "reserved": usage.total - usage.used - usage.free,
+                 "free": usage.free }
 
 class Stream(NilmApp):
     """Stream-specific operations"""
@@ -199,10 +106,10 @@ class Stream(NilmApp):
         layout parameter, just list streams that match the given path
         or layout.
 
-        If extent is not given, returns a list of lists containing
+        If extended is missing or zero, returns a list of lists
-        the path and layout: [ path, layout ]
+        containing the path and layout: [ path, layout ]
 
-        If extended is provided, returns a list of lists containing
+        If extended is true, returns a list of lists containing
         extended info: [ path, layout, extent_min, extent_max,
         total_rows, total_seconds ].  More data may be added.
         """
@@ -315,6 +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
@@ -439,6 +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
@@ -566,70 +479,14 @@ class Server(object):
 
     def json_error_page(self, status, message, traceback, version):
         """Return a custom error page in JSON so the client can parse it"""
-        errordata = { "status" : status,
+        return json_error_page(status, message, traceback, version,
-                      "message" : message,
+                               self.force_traceback)
-                      "traceback" : traceback }
-        # Don't send a traceback if the error was 400-499 (client's fault)
-        try:
-            code = int(status.split()[0])
-            if not self.force_traceback:
-                if code >= 400 and code <= 499:
-                    errordata["traceback"] = ""
-        except Exception: # pragma: no cover
-            pass
-        # Override the response type, which was previously set to text/html
-        cherrypy.serving.response.headers['Content-Type'] = (
-            "application/json;charset=utf-8" )
-        # Undo the HTML escaping that cherrypy's get_error_page function applies
-        # (cherrypy issue 1135)
-        for k, v in errordata.iteritems():
-            v = v.replace("&lt;","<")
-            v = v.replace("&gt;",">")
-            v = v.replace("&amp;","&")
-            errordata[k] = v
-        return json.dumps(errordata, separators=(',',':'))
 
     def start(self, blocking = False, event = None):
-
+        cherrypy_start(blocking, event, self.embedded)
-        if not self.embedded: # pragma: no cover
-            # Handle signals nicely
-            if hasattr(cherrypy.engine, "signal_handler"):
-                cherrypy.engine.signal_handler.subscribe()
-            if hasattr(cherrypy.engine, "console_control_handler"):
-                cherrypy.engine.console_control_handler.subscribe()
-
-        # Cherrypy stupidly calls os._exit(70) when it can't bind the
-        # port.  At least try to print a reasonable error and continue
-        # in this case, rather than just dying silently (as we would
-        # otherwise do in embedded mode)
-        real_exit = os._exit
-        def fake_exit(code): # pragma: no cover
-            if code == os.EX_SOFTWARE:
-                fprintf(sys.stderr, "error: CherryPy called os._exit!\n")
-            else:
-                real_exit(code)
-        os._exit = fake_exit
-        cherrypy.engine.start()
-        os._exit = real_exit
-
-        # Signal that the engine has started successfully
-        if event is not None:
-            event.set()
-
-        if blocking:
-            try:
-                cherrypy.engine.wait(cherrypy.engine.states.EXITING,
-                                     interval = 0.1, channel = 'main')
-            except (KeyboardInterrupt, IOError): # pragma: no cover
-                cherrypy.engine.log('Keyboard Interrupt: shutting down bus')
-                cherrypy.engine.exit()
-            except SystemExit: # pragma: no cover
-                cherrypy.engine.log('SystemExit raised: shutting down bus')
-                cherrypy.engine.exit()
-                raise
 
     def stop(self):
-        cherrypy.engine.exit()
+        cherrypy_stop()
 
 # Use a single global nilmdb.server.NilmDB and nilmdb.server.Server
 # instance since the database can only be opened once.  For this to

nilmdb/server/serverutil.py

@@ -0,0 +1,214 @@
+"""Miscellaneous decorators and other helpers for running a CherryPy
+server"""
+
+import cherrypy
+import sys
+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."""
+    # Set this to False to get better tracebacks from some requests
+    # (/stream/extract, /stream/intervals).
+    func._cp_config = { 'response.stream': True }
+    return func
+
+def response_type(content_type):
+    """Return a decorator-generating function that sets the
+    response type to the specified string."""
+    def wrapper(func, *args, **kwargs):
+        cherrypy.response.headers['Content-Type'] = content_type
+        return func(*args, **kwargs)
+    return decorator.decorator(wrapper)
+
+@decorator.decorator
+def workaround_cp_bug_1200(func, *args, **kwargs): # pragma: no cover
+    """Decorator to work around CherryPy bug #1200 in a response
+    generator.
+
+    Even if chunked responses are disabled, LookupError or
+    UnicodeError exceptions may still be swallowed by CherryPy due to
+    bug #1200.  This throws them as generic Exceptions instead so that
+    they make it through.
+    """
+    exc_info = None
+    try:
+        for val in func(*args, **kwargs):
+            yield val
+    except (LookupError, UnicodeError):
+        # Re-raise it, but maintain the original traceback
+        exc_info = sys.exc_info()
+        new_exc = Exception(exc_info[0].__name__ + ": " + str(exc_info[1]))
+        raise new_exc, None, exc_info[2]
+    finally:
+        del exc_info
+
+def exception_to_httperror(*expected):
+    """Return a decorator-generating function that catches expected
+    errors and throws a HTTPError describing it instead.
+
+        @exception_to_httperror(NilmDBError, ValueError)
+        def foo():
+            pass
+    """
+    def wrapper(func, *args, **kwargs):
+        exc_info = None
+        try:
+            return func(*args, **kwargs)
+        except expected:
+            # Re-raise it, but maintain the original traceback
+            exc_info = sys.exc_info()
+            new_exc = cherrypy.HTTPError("400 Bad Request", str(exc_info[1]))
+            raise new_exc, None, exc_info[2]
+        finally:
+            del exc_info
+    # We need to preserve the function's argspecs for CherryPy to
+    # handle argument errors correctly.  Decorator.decorator takes
+    # care of that.
+    return decorator.decorator(wrapper)
+
+# Custom CherryPy tools
+
+def CORS_allow(methods):
+    """This does several things:
+
+    Handles CORS preflight requests.
+    Adds Allow: header to all requests.
+    Raise 405 if request.method not in method.
+
+    It is similar to cherrypy.tools.allow, with the CORS stuff added.
+
+    Add this to CherryPy with:
+    cherrypy.tools.CORS_allow = cherrypy.Tool('on_start_resource', CORS_allow)
+    """
+    request = cherrypy.request.headers
+    response = cherrypy.response.headers
+
+    if not isinstance(methods, (tuple, list)): # pragma: no cover
+        methods = [ methods ]
+    methods = [ m.upper() for m in methods if m ]
+    if not methods: # pragma: no cover
+        methods = [ 'GET', 'HEAD' ]
+    elif 'GET' in methods and 'HEAD' not in methods: # pragma: no cover
+        methods.append('HEAD')
+    response['Allow'] = ', '.join(methods)
+
+    # Allow all origins
+    if 'Origin' in request:
+        response['Access-Control-Allow-Origin'] = request['Origin']
+
+    # If it's a CORS request, send response.
+    request_method = request.get("Access-Control-Request-Method", None)
+    request_headers = request.get("Access-Control-Request-Headers", None)
+    if (cherrypy.request.method == "OPTIONS" and
+        request_method and request_headers):
+        response['Access-Control-Allow-Headers'] = request_headers
+        response['Access-Control-Allow-Methods'] = ', '.join(methods)
+        # Try to stop further processing and return a 200 OK
+        cherrypy.response.status = "200 OK"
+        cherrypy.response.body = ""
+        cherrypy.request.handler = lambda: ""
+        return
+
+    # Reject methods that were not explicitly allowed
+    if cherrypy.request.method not in methods:
+        raise cherrypy.HTTPError(405)
+
+
+# Helper for json_in tool to process JSON data into normal request
+# parameters.
+def json_to_request_params(body):
+    cherrypy.lib.jsontools.json_processor(body)
+    if not isinstance(cherrypy.request.json, dict):
+        raise cherrypy.HTTPError(415)
+    cherrypy.request.params.update(cherrypy.request.json)
+
+# Used as an "error_page.default" handler
+def json_error_page(status, message, traceback, version,
+                    force_traceback = False):
+    """Return a custom error page in JSON so the client can parse it"""
+    errordata = { "status" : status,
+                  "message" : message,
+                  "traceback" : traceback }
+    # Don't send a traceback if the error was 400-499 (client's fault)
+    try:
+        code = int(status.split()[0])
+        if not force_traceback:
+            if code >= 400 and code <= 499:
+                errordata["traceback"] = ""
+    except Exception: # pragma: no cover
+        pass
+    # Override the response type, which was previously set to text/html
+    cherrypy.serving.response.headers['Content-Type'] = (
+        "application/json;charset=utf-8" )
+    # Undo the HTML escaping that cherrypy's get_error_page function applies
+    # (cherrypy issue 1135)
+    for k, v in errordata.iteritems():
+        v = v.replace("&lt;","<")
+        v = v.replace("&gt;",">")
+        v = v.replace("&amp;","&")
+        errordata[k] = v
+    return json.dumps(errordata, separators=(',',':'))
+
+# Start/stop CherryPy standalone server
+def cherrypy_start(blocking = False, event = False, embedded = False):
+    """Start the CherryPy server, handling errors and signals
+    somewhat gracefully."""
+
+    if not embedded: # pragma: no cover
+        # Handle signals nicely
+        if hasattr(cherrypy.engine, "signal_handler"):
+            cherrypy.engine.signal_handler.subscribe()
+        if hasattr(cherrypy.engine, "console_control_handler"):
+            cherrypy.engine.console_control_handler.subscribe()
+
+    # Cherrypy stupidly calls os._exit(70) when it can't bind the
+    # port.  At least try to print a reasonable error and continue
+    # in this case, rather than just dying silently (as we would
+    # otherwise do in embedded mode)
+    real_exit = os._exit
+    def fake_exit(code): # pragma: no cover
+        if code == os.EX_SOFTWARE:
+            fprintf(sys.stderr, "error: CherryPy called os._exit!\n")
+        else:
+            real_exit(code)
+    os._exit = fake_exit
+    cherrypy.engine.start()
+    os._exit = real_exit
+
+    # Signal that the engine has started successfully
+    if event is not None:
+        event.set()
+
+    if blocking:
+        try:
+            cherrypy.engine.wait(cherrypy.engine.states.EXITING,
+                                 interval = 0.1, channel = 'main')
+        except (KeyboardInterrupt, IOError): # pragma: no cover
+            cherrypy.engine.log('Keyboard Interrupt: shutting down bus')
+            cherrypy.engine.exit()
+        except SystemExit: # pragma: no cover
+            cherrypy.engine.log('SystemExit raised: shutting down bus')
+            cherrypy.engine.exit()
+            raise
+
+# Stop CherryPy server
+def cherrypy_stop():
+    cherrypy.engine.exit()

nilmdb/utils/__init__.py

@@ -14,3 +14,4 @@ import nilmdb.utils.iterator
 import nilmdb.utils.interval
 import nilmdb.utils.lock
 import nilmdb.utils.sort
+import nilmdb.utils.unicode

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)
@@ -104,3 +109,20 @@ def set_difference(a, b):
             b_interval = None
             if a_interval:
                 out_start = ts
+
+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/unicode.py

@@ -0,0 +1,29 @@
+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, text_type):
+        return u
+    try:
+        return u.encode()
+    except UnicodeEncodeError:
+        return u.encode("utf-8")
+
+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, text_type):
+        return s
+    try:
+        return s.decode()
+    except UnicodeDecodeError:
+        try:
+            return s.decode("utf-8")
+        except UnicodeDecodeError:
+            return s # best we can do

tests/data/timestamped

@@ -0,0 +1,8 @@
+-10000000000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+-100000000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+-100000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+-1000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+1 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+1000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+1000000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  
+1000000000 2.61246e+05  2.22735e+05  4.60340e+03  2.58221e+03  8.42804e+03  3.41890e+03  9.57898e+02  4.00585e+03  

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",
+        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

@@ -88,7 +88,7 @@ class TestCmdline(object):
                 sys.exit(0)
             except SystemExit as e:
                 exitcode = e.code
-        captured = outfile.getvalue()
+        captured = nilmdb.utils.unicode.decode(outfile.getvalue())
         self.captured = captured
         self.exitcode = exitcode
 
@@ -245,8 +245,10 @@ class TestCmdline(object):
         self.contain("Client version: " + nilmdb.__version__)
         self.contain("Server version: " + test_server.version)
         self.contain("Server database path")
-        self.contain("Server database size")
+        self.contain("Server disk space used by NilmDB")
-        self.contain("Server database free space")
+        self.contain("Server disk space used by other")
+        self.contain("Server disk space reserved")
+        self.contain("Server disk space free")
 
     def test_04_createlist(self):
         # Basic stream tests, like those in test_client.
@@ -473,6 +475,13 @@ class TestCmdline(object):
         # bad start time
         self.fail("insert -t -r 120 --start 'whatever' /newton/prep /dev/null")
 
+        # Test negative times
+        self.ok("insert --start @-10000000000 --end @1000000001 /newton/prep"
+                " tests/data/timestamped")
+        self.ok("extract -c /newton/prep --start min --end @1000000001")
+        self.match("8\n")
+        self.ok("remove /newton/prep --start min --end @1000000001")
+
     def test_07_detail_extended(self):
         # Just count the number of lines, it's probably fine
         self.ok("list --detail")

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"

tests/test_nilmdb.py

@@ -157,11 +157,14 @@ class TestServer(object):
 
     def test_server(self):
         # Make sure we can't force an exit, and test other 404 errors
-        for url in [ "/exit", "/", "/favicon.ico" ]:
+        for url in [ "/exit", "/favicon.ico" ]:
             with assert_raises(HTTPError) as e:
                 geturl(url)
             eq_(e.exception.code, 404)
 
+        # Root page
+        in_("This is NilmDB", geturl("/"))
+
         # Check version
         eq_(distutils.version.LooseVersion(getjson("/version")),
             distutils.version.LooseVersion(nilmdb.__version__))