Unknown fields

Unknown fields

A field the catalog does not cover is examined rather than dropped. Two things can be said about one without guessing, and they are kept apart.

Only hardware that uploads has a catalog. A station this machine reads over a cable hands over WeeWX fields directly, so it has no unknown fields and nothing here happens to it.

Derived

The field continues a series the catalog describes. If zz_ch1 and zz_ch2 are known, zz_ch3 follows from them: the hardware numbers its own channels, and the catalog supplies both ends of the series.

INFO user.ultimatepush.mapping: New field 'leafwetness_ch5' -> 'leafWet5' (group_percent),
continues leafwetness_ch, e.g. leafWet1

Taken automatically, unless the family's placement is a convention rather than a reading. Those wait for you, and the log gives the line to add:

INFO user.ultimatepush.mapping: New channel 'temp9f' would go to 'extraTemp9'.
Which sensor that is, and whether that field is free, only you know. Add
'temp9f = extraTemp9' under [[field_map_extensions]] to accept it.

Knowing where a channel belongs is not the same as knowing the field is free. A new WN34 channel would go to extraTemp, where a sensor configured two years ago may already have history, and two series in one column cannot be separated afterwards. A family with nowhere else to be placed, such as a laser rangefinder's depth or a lightning count, is taken without asking. See Field map.

Guessed

The name says what the reading is. Most of this hardware is consistent about it:

Pattern	Reading
rssi	signal strength, dB
_sig	signal quality, a count
batt	battery
_time	a timestamp
barom...in	pressure, inHg
rain...in, rain...piezo	rain, inches
mph	speed, mph
winddir...	direction, degrees
temp..., tf_..., soiltemp, thermo	temperature, °F
humidity, moisture, _hum	percent
pm1, pm4, pm10, pm25	concentration, µg/m³
co2, co	ppm
solarradiation, radiation	W/m²
uv	UV index
vpd	pressure, kPa
depth_ch, air_ch, thi_ch	distance, mm

Reported, not taken:

INFO user.ultimatepush.mapping: New field 'yearlyrainin' looks like group_rain
(name matches rain.*in$|rain.*piezo$), but it was only guessed. Left out.
Add it to field_map_extensions to keep it.

A guessed field goes to the protocol's own prefix plus the raw name: ecowitt_, wu_, ambient_, acurite_ and so on. Two protocols that send the same unrecognised name therefore land in different columns, which is the safe way round: a name nobody has identified is not known to mean the same thing in both.

Past the published channel count

Each maker publishes how many channels a sensor supports. A channel beyond that is real but not routine, so it is reported rather than derived:

soilmoisture17   soilMoist17   channel 17, past the 16 a WH51 is said to support

Either the published figures have moved on, or something else is going on. Both are worth a look before the reading lands in a column.

Nothing can be said

INFO user.ultimatepush.mapping: No idea what 'wizzlefrob' is. Left out.

Please report it. See Reporting a new sensor.

The setting

[UltimatePush]
    infer_unknown = series

Value	Effect
off	Nothing is taken. Everything unknown is logged and dropped.
series	Derived fields are taken, unless their placement is a convention. Guesses are logged. Default.
all	Guesses are taken too, under the protocol's prefix plus the raw name.

all gets you the reading sooner, at the risk of a unit nobody checked. It is the right setting while working out what a station sends, which is why python -m user.ultimatepush uses it.

Whatever the setting, a field is reported once per run. Restart the driver to see the messages again.