You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It really seems like there ought to be a clear distinction between attributes, properties, and functions on RTD. The current settings for sphinx don't accomplish this.
In particular, drift_state, total_samples, samples_since_reset are common across all detectors, but to get them to display as inherited using the :inherit-members: option in sphinx, it was necessary to set them up as properties. Only drift_state really needs the validation enabled by the @property decorator, though.
The choice was between two evils: 2 of the 3 appearing in Attributes, redundantly specified in the child classes; or all 3 appearing as properties, documented only in the parents in drift_detector.
If there's a way around this that avoids static .rst files, that'd be fantastic. Otherwise, we may someday have to bite the bullet of maintaining those.
One option for ordering them in a particular way is to use the "member-order":"bysource" option in conf.py, so they'll be ordered according to the source code (seems easy to mess up!), but this still keeps the inherited properties at the bottom, so it's not a big potential improvement over groupwise at the moment.
The text was updated successfully, but these errors were encountered:
It really seems like there ought to be a clear distinction between attributes, properties, and functions on RTD. The current settings for sphinx don't accomplish this.
In particular,
drift_state, total_samples, samples_since_reset
are common across all detectors, but to get them to display as inherited using the :inherit-members: option in sphinx, it was necessary to set them up as properties. Onlydrift_state
really needs the validation enabled by the@property
decorator, though.The choice was between two evils: 2 of the 3 appearing in Attributes, redundantly specified in the child classes; or all 3 appearing as properties, documented only in the parents in
drift_detector
.If there's a way around this that avoids static
.rst
files, that'd be fantastic. Otherwise, we may someday have to bite the bullet of maintaining those.One option for ordering them in a particular way is to use the
"member-order":"bysource"
option inconf.py
, so they'll be ordered according to the source code (seems easy to mess up!), but this still keeps the inherited properties at the bottom, so it's not a big potential improvement overgroupwise
at the moment.The text was updated successfully, but these errors were encountered: