I’ve noticed that when using very large files with the MergeView editor, the computed diff between the files is incorrect.
More specifically for my case, I have 2 ~1500+ loc files that have just a couple changes between them (not more than 40-50 loc) spread out throughout the whole document (small diffs at the beginning, middle, and end of the document).
Here’s the diff that I’m expecting to see when using a tool like diffchecker.com.
And this is how it’s rendered in CM using the MergeView. The whole document is basically flagged as one large diff.
Here’s a Gist with repro documents for testing and a playground configured with MergeView. Couldn’t generate a proper share link since the url is too long when using my repro documents.
Computing the diff for such large (and different) files is too expensive to do interactively, so the editor falls back to an overapproximation of the diff.
If people are allowed to edit the document, it seems that having a slow (and this is super-linear, so it gets very slow quite quickly) UI update cycle would make it unusable pretty quickly.
If users aren’t allowed to edit, maybe just statically rendering the text is easier than having it in a CodeMirror editor.
My use case is very similar to GitHub’s diff mode in PRs. I use CM to look at the version of a source file or compare the source of a file across two different versions. The document is static to the user and is indeed initialized with EditorState.readOnly.of(true) and EditorView.editable.of(false).
If I understand correctly there’s no way of making CM render the content statically and get my desired outcome? Should I be looking at a different framework like prism for my use cases?
Computing the diff for such large (and different) files is too expensive to do interactively, so the editor falls back to an overapproximation of the diff.
Could you go a bit more into detail in what is considered too expensive (file size, word count, etc) so I can get an idea of when the editor is expected to fallback to the approximation?
Awesome! I’ve put together a repo to try it out and it’s working as expected when setting the scanLimit (preview link).
Right now, I’m running the dependency off the main branch from GitHub. Once you cut a new release and make it official, I will update my package to the npm version.
