examples: Add dup filtering to mqtt_relay

Keep information about the previous value sent.  If it's been 8
seconds, or new value is different (ignorning keys like snr and
frequency), then send it.  Otherwise, just don't.  This causes bursts
of e.g. 4 transmissions to result in one MQTT message, on the theory
that the 4 transmissions are not actually 4 messags, but a strategy to
transmit one message more reliably.

Define a new configuration option to enable duplicate filtering, and
default it to True.

Author: gdt

examples/rtl_433_mqtt_relay.py

@@ -20,10 +20,13 @@
 from __future__ import print_function
 from __future__ import with_statement
 
-import socket
 import json
+import socket
+import time
+
 import paho.mqtt.client as mqtt
 
+
 # The config class represents a config object.  The constructor takes
 # an optional pathname, and will switch on the suffix (.yaml for now)
 # and read a dictionary.
@@ -42,6 +45,7 @@ class rtlconfig(object):
         'MQTT_PASSWORD': None,
         'MQTT_TLS': False,
         'MQTT_PREFIX': "sensor/rtl_433",
+        'MQTT_DEDUP': True,
         'MQTT_INDIVIDUAL_TOPICS': True,
         'MQTT_JSON_TOPIC': True,
     }
@@ -68,9 +72,75 @@ def __init__(self, f=None):
     def __getitem__(self, k):
         return self.c[k]
 
+# A dedup class object supports deduping a stream of reports by
+# answering if a report is interesting relative to the history.
+# While more complicated deduping is allowed by the interface, for now
+# it is very simple, keeping track of only the previous interesting object.
+# For now, we more or less require that all reports have the same keys.
+# \todo Consider a cache with several entries.
+class dedup(object):
+
+    def __init__(self):
+        # Make this long enough to skip repeats, but allow messages
+        # every 10s to come through.
+        self.duration = 5
+        self.boring_keys = ('time', 'freq', 'rssi', 'snr', 'noise')
+        # Initialize storage for what was last sent.
+        (self.last_report, self.last_now) = (None, None)
+    
+    def send_store(self, report, n):
+        (self.last_report, self.last_now) = (report, n)
+        return True
+
+    # Return True if j1 and j2 are the same, except for boring_keys.
+    def equiv(self, j1, j2):
+        for (k, v) in j1.items():
+            # If in boring, we don't care.
+            if k not in self.boring_keys:
+                # If in j1 and not j2, they are different.
+                if k not in j2:
+                    return False
+                if j1[k] != j2[k]:
+                    return False
+        # If the lengths are different, they must be different.
+        if len(j1) != len(j2): 
+            return False
+
+        # If we get here, then the lengths are the same, and all
+        # non-boring keys in j1 exist in j2, and have the same value.
+        # It could be that j2 is missing a boring key and also has a
+        # new non-boring key, but boring keys in particular should not
+        # be variable.
+        return True
+
+    # report is a python dictionary
+    def is_interesting(self, report):
+        n = time.time()
+
+        # If previous interesting is empty (or troubled), accept this
+        # one.
+        if self.last_report is None or self.last_now is None:
+            # print("interesting: no previous")
+            return self.send_store(report, n)
+
+        # If previous one was too long ago, accept this one.
+        if n - self.last_now > self.duration:
+            # print("interesting: time")
+            return self.send_store(report, n)
+
+        if not self.equiv(self.last_report, report):
+            # print("interesting: different")
+            return self.send_store(report, n)
+
+        return False
+
 # Create a config object, defaults modified by the config file if present.
 c = rtlconfig("rtl_433_mqtt_relay.yaml")
 
+# Create a dedup object for later use, even if it's configure off.
+d = dedup()
+
+
 def mqtt_connect(client, userdata, flags, rc):
     """Handle MQTT connection callback."""
     print("MQTT connected: " + mqtt.connack_string(rc))
@@ -100,6 +170,18 @@ def sanitize(text):
 def publish_sensor_to_mqtt(mqttc, data, line):
     """Publish rtl_433 sensor data to MQTT."""
 
+    debug = False         # \todo Hoist to program-wide debug control.
+
+    if c['MQTT_DEDUP']:
+        # If this data is not novel relative to recent data, just skip it.
+        # Otherwise, send it via MQTT.
+        if not d.is_interesting(data):
+            if debug:
+                print("not interesting: %s" % (line))
+                return
+            if debug:
+                print("INTERESTING: %s" % (line))
+
     # Construct a topic from the information that identifies which
     # device this frame is from.
     # NB: id is only used if channel is not present.