search

ponder-lab / GitHub-Issue-Classifier

Python script to mine for GitHub issues + comments and classify them.
MIT License
6 stars 0 forks source link

Do not process issue descriptions #47

Closed khatchad closed 3 years ago

khatchad commented 3 years ago

Only process (classify) issue comments. Please ignore description.

y3pio commented 3 years ago

Just to clarify, currently a sample commentURL is as follows: "https://github.com/q-optimize/c3/issues/74"

But it should instead be: "https://github.com/q-optimize/c3/issues/74#issue-835058056" which anchors us to the specific comment for the result that we are looking at? (in the case of this issue, there's only 1 comment)

khatchad commented 3 years ago

Just to clarify, currently a sample commentURL is as follows: "https://github.com/q-optimize/c3/issues/74"

Yes.

But it should instead be: "https://github.com/q-optimize/c3/issues/74#issue-835058056" which anchors us to the specific comment for the result that we are looking at?

Yes.

(in the case of this issue, there's only 1 comment)

If that is the case, there should still be an anchor. I am just going off what I expect to see per CSV column name. My assumption is that you wouldn't deliberately have two columns with the same data.

y3pio commented 3 years ago

Yup, agreed. I will look into this to see if the API provides any way to append comment ID's to the end of the URL to anchor us to the specific comment.

khatchad commented 3 years ago

I would think that this information is given in the response JSON.

khatchad commented 3 years ago

https://docs.github.com/en/rest/reference/issues#comments

y3pio commented 3 years ago

Hi Professor, it seems like I'm already anchoring the comments. I think for issues with just one comment, it somehow shows up as the same URL as the issue URL. Per their API docs, the html_url field that I am using for the column should have an #issue-12345 pattern, but however the API seems to just return the main issue URL without the anchor. Below is the API call for the issue above that I mentioned:

https://api.github.com/repos/q-optimize/c3/issues/74

As you can see the html_url field does not contain the #issue-... anchor as the docs indicate. But for the comments found in the comments_url API call, I believe the tool is anchoring them accordingly.

Please let me know if my explanation is clear, happy to walk through this over a call if needed.

khatchad commented 3 years ago

Hi Professor, it seems like I'm already anchoring the comments.

At a quick glance, I didn't see any such anchoring. The ones I saw did not include anchors. Do you have any counter evidence?

I think for issues with just one comment, it somehow shows up as the same URL as the issue URL.

But not all issues will have only one comment. If this is the case, we should have at least one example of an anchored URL, right? Another examples of evidence would be the location in the code where this field is being populated.

Per their API docs, the html_url field that I am using for the column should have an #issue-12345 pattern, but however the API seems to just return the main issue URL without the anchor.

In all cases or just cases where there is only one comment?

Below is the API call for the issue above that I mentioned:

https://api.github.com/repos/q-optimize/c3/issues/74

There are no comments in this issue. Do we process issues with no comments?

As you can see the html_url field does not contain the #issue-... anchor as the docs indicate.

That's because it doesn't have any comments. In other words, https://api.github.com/repos/q-optimize/c3/issues/74/comments gives a empty array.

But for the comments found in the comments_url API call, I believe the tool is anchoring them accordingly.

My impression is that the dataset is a list of issue comments. Why are we examining an issue that has no comments? Or, do you treat issue descriptions as comments?

Please let me know if my explanation is clear, happy to walk through this over a call if needed.

I think it makes me more confused :). Sure, let's do a call. Please send me a calendar invite when you have a chance. Thanks @y3pio.