search

apertium / phenny

This is a port of phenny, a Python IRC bot, to Python3. This specific version, called begiak, is a fork of the Wadsworth version, used by the apertium project.
http://wiki.apertium.org/wiki/Begiak
Other
16 stars 42 forks source link

.awikstats misc issues #486

Open jonorthwash opened 5 years ago

jonorthwash commented 5 years ago

I ran .awikstats update kaz-kaa and this is what it did.

Issues I see:

sushain97 commented 5 years ago
  • The username francis.m..tyers is not a github username, nor is it even a valid part of one. It should be @ftyers, I think.

This is as designed but a bit unfortunate. The username's are GitHub's translation of a GH user to an SVN user. https://github.com/apertium/apertium-stats-service/issues/52 should be able to modify the behavior.

  • It added the incorrect rlx rlx rules instead of updating rlx rules.
  • It added dix stems instead of updating (/renaming) stems.

This sounds like bugs in this repo.

This one is definitely by design. Not sure how to generate the latter URL since I'm not sure what the 593d5b5a1eb858097a6f44040ef3cb03 is referring to....

jonorthwash commented 5 years ago
  • The username francis.m..tyers is not a github username, nor is it even a valid part of one. It should be @ftyers, I think.

This is as designed but a bit unfortunate. The username's are GitHub's translation of a GH user to an SVN user. apertium/apertium-stats-service#52 should be able to modify the behavior.

With the two dots also?

  • It added the incorrect rlx rlx rules instead of updating rlx rules.
  • It added dix stems instead of updating (/renaming) stems.

This sounds like bugs in this repo.

By "this repo" do you mean phenny or apertium-kaz-kaa? If the latter, what's the relationship between what's in the repo and what it prints for rule counts?

This one is definitely by design. Not sure how to generate the latter URL since I'm not sure what the 593d5b5a1eb858097a6f44040ef3cb03 is referring to....

If you click on it, it anchors to the right file. It's also what's linked to for the commit under the file's history. I think even loading that page of diffs for the commit and not having the anchor to the specific file would be preferable to what we have now, since it's in the context of clicking on the commit hash, not the filename.

sushain97 commented 5 years ago
  • The username francis.m..tyers is not a github username, nor is it even a valid part of one. It should be @ftyers, I think.

This is as designed but a bit unfortunate. The username's are GitHub's translation of a GH user to an SVN user. apertium/apertium-stats-service#52 should be able to modify the behavior.

With the two dots also?

Yep.

  • It added the incorrect rlx rlx rules instead of updating rlx rules.
  • It added dix stems instead of updating (/renaming) stems.

This sounds like bugs in this repo.

By "this repo" do you mean phenny or apertium-kaz-kaa? If the latter, what's the relationship between what's in the repo and what it prints for rule counts?

phenny. This repo doesn't do any of the calculation, it just gets the numbers from stats-service and updates the wiki. Numbers: https://apertium.projectjj.com/stats-service/apertium-kaz-kaa.

This one is definitely by design. Not sure how to generate the latter URL since I'm not sure what the 593d5b5a1eb858097a6f44040ef3cb03 is referring to....

If you click on it, it anchors to the right file. It's also what's linked to for the commit under the file's history. I think even loading that page of diffs for the commit and not having the anchor to the specific file would be preferable to what we have now, since it's in the context of clicking on the commit hash, not the filename.

Hmm, yeah, I see what the point of the anchor is. Unfortunately, I don't actually know what that value is and how to generate it programatically. As for going to the diff instead of the file, I think we were replicating the old behavior. However, the diff is doable as well. That would be a change in phenny (not a large one I think).

jonorthwash commented 5 years ago

This is as designed but a bit unfortunate. The username's are GitHub's translation of a GH user to an SVN user. apertium/apertium-stats-service#52 should be able to modify the behavior.

At http://wiki.apertium.org/wiki/Apertium-kaz-kir/stats I'm getting myself as both "jonorthwash" and "jonathan.north.washington". Are you sure this is as designed?

jonorthwash commented 5 years ago

Some action items so far:

jonorthwash commented 5 years ago

Also, it looks like transfer macros aren't being counted correctly. The kaz-kir stats page reports 0 kaz-kir t1x macros, but kaz-kir.t1x has 6.

sushain97 commented 5 years ago

At http://wiki.apertium.org/wiki/Apertium-kaz-kir/stats I'm getting myself as both "jonorthwash" and "jonathan.north.washington". Are you sure this is as designed?

Yep. This seems attributable to different emails being used which appears is what GH uses to create the SVN authors (at least partially)? It's a bit unclear. Note the two emails below for you.

$ git log --pretty=format:"%ae" | sort | uniq | cut -d '@' -f 1
ftyers
ilnar
itang1
jonathan.n.washington
jonathan.north.washington
kakubay
kantoro90
sushain
unhammer

Also, it looks like transfer macros aren't being counted correctly. The kaz-kir stats page reports 0 kaz-kir t1x macros, but kaz-kir.t1x has 6.

Hmm, this one sounds like a bug. Does this look like the right way to count them?

https://github.com/apertium/apertium-stats-service/blob/master/src/stats/xml.rs#L95

IIRC, I copied whatever the old Python scripts did.

jonorthwash commented 5 years ago

Does this look like the right way to count them?

No, it should be counting <def-macro ...>...</def-macro> blocks.

sushain97 commented 5 years ago

Ah... I wonder why I wrote it the other way. Fixed: https://github.com/apertium/apertium-stats-service/commit/ac4cbacb34f0eed65c6920b8e84897f8229794ae.

I'll deploy it in a bit.

sushain97 commented 5 years ago

@jonorthwash look any better?

jonorthwash commented 5 years ago

The raw data at https://apertium.projectjj.com/stats-service/apertium-kaz-kir looks good now; I'll have to check the output of .awikstats later.