Posted by andreas.riegg
Config options for devices allow setting of parameters like I2C slave address. (Currently) they have to be in the right order as they are used in __init__ of the device class. Sometimes lets say I want to change the 3rd parameter. In order to do this its also necessary to provide the 1st and 2nd parameter even if they can keep their defualt values. One possibility would be to copy the parameter=value scheme from __init__ into config. But that would be heavy additionally parsing effort.
A simple solution would be to allow e.g. "*" as placeholders for parameters that can keep their default values from __init__. The number/sequence of parameters will still have to be the correct one (trailing ones not nedded to be changed at the end can be ommitted), but errors arising from wrong hard-coded default values in config can be avoided.
Example
mydevice = DEVICECLASS * * 13
On implementation side, the code that scans config parameters could just pass "None" on all parameter positions that can keep default values so that the __init__ values provide the missing value again.
Posted by trouch
I would like a more explaining example with a specific device class for instance.
Posted by trouch
I checked VCNL, and I understand the problem... First, are all arguments supposed to be tweak ? Then, instead of using wildcard, we could use key/value pair : light0 = VCNL4000 current=40 or light0 = VCNL4000(current=40)
but I have to check how my config parser will handle multiple '=' in a line
Posted by andreas.riegg
Some more examples:
TMPx75:
mytmp = TMP75 * 10
Set the resolution to 10 bit but keeping the default slave address
If we (may) add setting the upper and lower alert/thermostat registers of them as parameter 3 and 4:
mytmpwithalerts = TMP75 * * 10 40
Set the temps to 10 deg low and 40 deg high but keep the default slave address and default resulution
TSL2561 with high gain:
mytslhighgain = TSL2561T * * 16
Posted by andreas.riegg
or
light0 = VCNL4000 current:40 as the ":" may be free to simplify parsing
Posted by trouch
Well I tried the two options :
1) passing None to a function override defaults. So defaults values has to be handled manually in the code instead of __init__ declaration with wildcard support.
2) key/value pair args cannot be mixed with simple value args in config file. We always need to precise arg names.
giving this, I think opt. 2 is clearer and more consistent, so I committed in r976. I let you try with TMPx75 or VCNL before closing the issue.
Posted by andreas.riegg
I was able to test the new config option with tsl and vcnl and can confirm it working perfect ok. Thanks for updating. I was also able to test devices-monitor with all my sensors attached. See attached screenshots. This also confirms the distance part of the vcnl driver mapping correct for devices-monitor.
During my test I moved the vcnl adapter to another place on my breadboard and .... what a wonder .... it started working. Assume the breakout has a bad smd soldered chip that re-got the connection by moving. Not very solid as I moved the beakout also sometimes before and this did not help, but as long as it works now .... Anyway, that allowed me to final test also the distance part of the vcnl driver and tweak the distance lookup table a bit so that I can confirm now also the distance part of the vcnl driver to be final for now and tested. As calibration seems to be necessary from time to time I added a return value to the calibrate method that gives the current calibration offset. This way from a macro someone can do local or remote calibration from time to time and see the result. Don't think this should be done via REST so I did not add a REST mapping for now. The final vcnl is also attached.
Files attached
Posted by trouch
Nice ! Thanks for feedback, and happy your VCNL is finally working. I close the issue, and I'll update the VCNL driver later today.
