‌‌‌‌‌‌‌Once this PR is merged, does this issue to some extent no longer exist?

#1316

Which issue do you mean? your link seems to be a PR instead of an issue.

PS. I just found some bugs in current implementation, please wait a sec.

Oh, you meant PR #1316 somehow fixes #1288 , I haven't tried that yet.

Does that work with cached responses? It's somehow difficult to reproduce that malformed response to test that issue.

I'm a little bit lost in current test framework, I'll leave the task of adding test cases to you if you don't mind.

@yetone I saw your fix in #1318, theoretically it avoids creating nonsense folders/files. But I still believe tree-sitter based parsing is a more robust way, as it provides a "best guess" when the code block wrappers are not properly matched.
For example, in issue #1288 , this PR will not only stop the unwanted file/folder creation, but also make the last 4 code blocks applicable.

Do you want me to make a similar change to extract_cursor_planning_code_snippets_map ?

@yetone I saw your fix in #1318, theoretically it avoids creating nonsense folders/files. But I still believe tree-sitter based parsing is a more robust way, as it provides a "best guess" when the code block wrappers are not properly matched. For example, in issue #1288 , this PR will not only stop the unwanted file/folder creation, but also make the last 4 code blocks applicable.

Do you want me to make a similar change to extract_cursor_planning_code_snippets_map ?

That would be better!

Just a reminder, I just changed the start_line and end_line of the codeblock from 0-indexed to 1-indexed in this PR. You also need to make corresponding changes in this PR.

https://github.com/yetone/avante.nvim/pull/1335/files#diff-ddcb4139fbc53f068a08051ae32665c76507f66698fe427fcfdd384a36dce2fcL788-R873

Just a reminder, I just changed the start_line and end_line of the codeblock from 0-indexed to 1-indexed in this PR. You also need to make corresponding changes in this PR.

https://github.com/yetone/avante.nvim/pull/1335/files#diff-ddcb4139fbc53f068a08051ae32665c76507f66698fe427fcfdd384a36dce2fcL788-R873

I've aligned my code to your change, but that change seems to cause new problems for me with planning turned on (another story).

Basically I think I'm done, with extract_cursor_planning_code_snippets_map and parse_codeblocks switched to a tree-sitter-based solution.

And I've tested on my cached malformed response, it works well no matter planning mode is on or off.

I've aligned my code to your change, but that change seems to cause new problems for me with planning turned on (another story).

Could you please elaborate on this issue?

I've aligned my code to your change, but that change seems to cause new problems for me with planning turned on (another story).

Could you please elaborate on this issue?

Never mind, it's fixed by your later commits.