How to clean dependabot PRs when body is used as merge message?

[tomdavidson]

Hello, I have kodiak and dependabot. Its fairly slick - thank you. On catch is that dependabot has PR bodies with lots of info and markup that makes my git log kinda messy. Here's a gist as an example of what shows in the commit log.

My full kodiak.toml is in this gist, but I am stripping the html form the comments:

[merge.message]
# works with squash commits and semeantic PR.
title = "pull_request_title"
body = "pull_request_body"
strip_html_comments = true

The informative dependabot PR body message is helpful ... just not right for the merge message. Can we have different messages per user? Any suggestions?

[sbdchd]

Sorry for the delay! Do you think the cut_body_after and cut_body_before params could be of use here?

I think there are some html comments in dependabot's pr descriptions that can be used for the markers

https://kodiakhq.com/docs/config-reference#mergemessagecut_body_before

[tomdavidson]

@sbdchd thanks. Yes, That is one path, but I am using those with our own PR templates. I think I could add some string from dependabot PR and add it to the PR so I can use the cut_body_* with both. Maybe different merge settings for different users is worth looking into?

[sbdchd]

related discussion from when cut_body_* was originally added: #550

[tomdavidson]

Ha nice call back to dependabot in #550! I think the cut_body_ lists could actually be more relevant for the itches I'm scratchin than different merge settings based on username - and I'm assuming it would be simpler and not necessarily mutually exclusive with future per user setting.

How about kodiak function treats the setting as a list, even if its a list of one, converting the str to a list for backward compatibility. No need to match up or have equal size lists for _after and _before, treat each element as logical OR and greedy - the largest cut wins ?

cut_body_before = ["a", "b", "c"]
cut_body_after = "wyoming"

# I dont know py, but something like this looks like it could shoehorn well
cut_index = functools.reduce(lambda a, b: a if a > b else b, cut_body_before_list))

[sbdchd]

I think people might be surprised by largest cut wins, maybe it would be better to have another setting:

cut_body: list[tuple[str, str]]

so then multiple start/end pairs could be passed:

cut_body = [["start_a", "end_a"], ["start_b", "end_b"]]

[tomdavidson]

It would be hard to be surprised by the tuples.

[chdsbd]

Should we open a new issue to track a feature request for this @tomdavidson?