Closed d-chambers closed 4 years ago
@sboltz mind taking a look at this? I know df_to_inventory
is a but weird with so many nested functions but otherwise should be straightforward.
Sure, I can take a look. Thanks for implementing it.
On May 28, 2020, at 18:21, Derrick notifications@github.com wrote:
@sboltz mind taking a look at this? I know df_to_inventory is a but weird with so many nested functions but otherwise should be straightforward.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.
Merging #182 into master will decrease coverage by
0.08%
. The diff coverage is97.90%
.
@@ Coverage Diff @@
## master #182 +/- ##
==========================================
- Coverage 97.32% 97.23% -0.09%
==========================================
Files 37 37
Lines 4069 4159 +90
Branches 647 661 +14
==========================================
+ Hits 3960 4044 +84
- Misses 52 54 +2
- Partials 57 61 +4
Flag | Coverage Δ | |
---|---|---|
#unittests | 97.23% <97.90%> (-0.09%) |
:arrow_down: |
Impacted Files | Coverage Δ | |
---|---|---|
obsplus/utils/stations.py | 96.66% <97.41%> (-3.34%) |
:arrow_down: |
obsplus/constants.py | 100.00% <100.00%> (ø) |
|
obsplus/exceptions.py | 100.00% <100.00%> (ø) |
|
obsplus/utils/docs.py | 100.00% <100.00%> (ø) |
|
obsplus/utils/misc.py | 98.35% <100.00%> (+0.08%) |
:arrow_up: |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update d3d31d6...23968c1. Read the comment docs.
No problem, it seemed like the right thing to do.
Thanks @sboltz, all good points. I think I addressed all of them and I also couldn't resist the urge to refactor a little more. I will give it one more look-over tomorrow.
I should note this has one small breaking change: For the datalogger_keys and sensor_keys columns, obsplus now expects json strings or python objects. Separating the keys with __ was kinda weird, and a mistake in my opinion.
I should note this has one small breaking change: For the datalogger_keys and sensor_keys columns, obsplus now expects json strings or python objects. Separating the keys with __ was kinda weird, and a mistake in my opinion.
I agree that that was a kind of weird way to do it (though convenient when you are too lazy to figure out the 'right' way to do it). I think taking that out makes sense.
In the face of ambiguity, refuse the temptation to guess.
I succumbed to the temperation to guess. It should probably raise an error if both NRL and station client are specified.
This PR adds the option to use a
get_stations_kwargs
row in a dataframe passed toobsplus.utils.stations.df_to_inventory
in order to download an existing channels client with the default ObsPy fdsn client or one provided to the function.For context, it is nice to keep a simple csv of channels and be able to edit it and regenerate an inventory rather than work with the highly nested inventory object directly. This adds just one more optional column that should make it easier to get the correct response for channels registered with an FDSN provider.