RTD/doc search results call Julia functions Python functions instead #11527
Comments
|
Probably a Sphinx default that we fail to override. Fix will probably be in JuliaDoc. Research time! |
|
Yes, the Julia domain inherits directly from the Python domain, with some overrides. I think this only affects search results. |
pao
changed the title
|
I have no idea where this can be coming from--the only string reference to "Python" in the domain we do in fact replace, the default domain is set to "jl" in conf.py, and there is no translation done (language is set to default which is |
|
This only happens for the "latest" doc. Is there any difference between the latest and the 0.3 version? |
|
And the Julia-CN doc search is also correct. |
|
Except the search index might still be getting a translation? |
|
Hrm...that's weird. I at least confirmed that the translation files aren't at fault. This looks like something thinks this reference is cross-domain (otherwise it would just be "function" according to the documentation of conf.py). |
|
Changes in JuliaDoc between version used on release-0.3 and master: JuliaLang/JuliaDoc@0dd9903...012be05 It doesn't appear that much has been done to conf.py that hasn't been backported, either, just some changes to the sidebars. Nothing that should be influencing this. |
|
So both of them are generated nightly as well? |
|
They are both generated on a branch push (there's a webook that GitHub issues.) |
|
It appears to be OK now??? |
|
Did anyone do anything and can this be closed now? |
|
Didn't do a darn thing, but confirm fixed. I assume this is somehow RTD's doing. |
|
@pao Sorry to bother you again but the problem seems to be back ....... |
|
Still only for |
|
Ping. @pao ? The search result is still showing "Python Function" now... |
|
We literally don't control this in any way, as proven by the fact it changed, then changed back, without touching anything. Perhaps we should bring it up with RTD. |
|
Do they have a tracker? |
|
RTD hasn't made any changes that would have affected this, I do see what looks like a corrupted build env on one of the servers however. This could result in transient failures like this -- though still not sure why the failure case generates invalid indexed titles. |
|
Oh, and with maintainer status on the project, you can trigger a environment wipe on the project's version page: |
|
cc @nolta @StefanKarpinski would one of you be able to trigger the environment wipe for |
|
Ok, added and wiped. |
|
That does not appear to have fixed it. |
|
Okay, after a deep dark rabbit hole, the domain in your docs now has the proper prefix on the latest build :) This is using the same server that was problematic before, so hopefully this makes the builds consistent. |
|
Deep and dark, indeed. Thanks for taking a look, and for the fix. |
|
@agjohnson Sorry to bother you again, but the original issue reported here ("Python function" showing up in search results) is still present; are we waiting for the fix to be pushed to RTD in production? Or did I misunderstand what all got fixed (was it just the environment installation problem)? |
|
@pao It was fixed, but seems one of the other build servers we are decommissioning has the outdated readthedocs-sphinx-ext 0.5.3 package, I'll update all of the build servers this morning. |
|
Got it, thanks for the clarification. |
|
I don't see "Python function" any more, looks fixed now. |
|
Fixed here too. Hopefully it won't come back anymore..... |
|
@agjohnson @pao It comes back for me again right now and IIRC at least once in the past week. |
|
Seems to be transient (most recent works for me right now)--maybe one of their build servers is/was still not updated? |
|
We just rolled all new build servers, so there may have been a point where the old extension was being pulled in. Definitely raise the issue on our tracker if it comes up again. |
|
Will do, thanks. |
|
@agjohnson It is "Python function" at this moment again.... (and IIRC was correct sometime yesterday) |
|
Reported at readthedocs/readthedocs.org#1382 |
|
Depending on when it was built, this could easily have something to do with #11906 ? |
|
It's now back to saying Julia function, but it's been off and on this entire time. |
|
@tkelman, since none of that has touched the definitions in JuliaDoc, we shouldn't be able to break it from within the doc sources themselves, unless we explicitly generated rst which specifies that they are Python functions. Julia is the default type. |
yuyichao
removed
the
kind:upstream
|
Doesn't seems seems to be referenced here so JuliaLang/JuliaDoc#16 |
|
This is fixed and superseded by the new doc system. |
I wanted to read the documentation for
Base.copy, and to my surprise the labels reference them as Python functions.Did someone do a bulk overwrite on the docs? Here's the link and a screenshot
http://julia.readthedocs.org/en/latest/search/?q=copy&check_keywords=yes&area=default
The text was updated successfully, but these errors were encountered: