[Refactor][Bitbucket_Server] Speed up PR collector/extractor

[sstojak1]

What and why to refactor

What are you trying to refactor? Why should it be refactored now?
The pr_collector for Bitbucket Server consistently adds the same data to the RAW_PULL_REQUEST_TABLE after each run.
Consequently, the extractApiPullRequests process slows down because it has to sift through all the records in the raw table, including duplicates.
For instance, if a repository has 1000 pull requests, after 10 job runs, the raw table will contain 10,000 rows, and extractApiPullRequests will have to process each of these records.

Was there a need to have all those history raw API data imports and because of that delete is not a feasible option?

Describe the solution you'd like

How to refactor?

1. Perhaps we could go with deleting all records from the raw table and importing PR again with each job run. This would prevent duplicates and avoid slowing down extractApiPullRequests task.
2. Check how Bitbucket plugin is doing it and maybe reuse the logic if it's better?

Related issues

Please link any other

Additional context

Add any other context or screenshots about the feature request here.

How to recreate:
Run Collect Data for Bitbucket Server more than once and observe the size of _raw_bitbucket_server_api_pull_requests table.

[Startrekzky]

Hi @sstojak1 , it's a valid refactore, but may I know why the same data keeps been adding to the RAW_PULL_REQUEST_TABLE after each run?

[sstojak1]

Hi @Startrekzky I would say that this is because BB Server API doesn't support date query parameter. So you cannot fetch PRs that were updated/created after last import job.

Because of that Devlake is importing all PRs during each job run (check how BB Cloud is not doing that since date query parameter is supported for that tool).

[klesh]

Based on the information you provided, it seems the collector would benefit from using the simpler ApiCollector since it would purge related records from the raw data table before saving new PR information. This might be more efficient compared to the StatefulApiCollector in this context.

[sstojak1]

@klesh That might work! Is there a way to run Devlake locally in debug mode? I'd like to go through the ApiCollector impl to understand its impact on the rest of the steps for importing BB server data...

[klesh]

Yes, sure. You may follow this guide.

In case you wanna execute specific subtasks, you may go to the backend folder and run sth like the following:

go run plugins/jira/jira.go -c 2 -b 8 -t "extractWorklogs"

remember to change the plugin name and arguments accordingly.