Conversation
|
Regarding startup time, I tested startup time on three larger projects:
They all loaded quickly. Only a 3800+ lines tex file of the research proposal took some two seconds to load. But that was also true for the current master branch. |
|
I see virtually no difference when typesetting my (admittedly too long) 329 page PhD thesis. The new commits don't seem to have any impact either way. |
|
I checked on my biggest file (the manual for |
|
No measurable difference for me (using |
|
+1 No appreciable difference for me between the feature branch and master on a ~300 page book with multiple chapters in separate files |
|
Thank you all for the quick feedback! I see no reason to merge this, then. |
Performance evaluation for the work on PR lervag#2558 gave evidence that disabling the directory scan method for the main file detection due to performance concerns appears to be unjustified. Remove this option. Note that recently g:vimtex_disable_recursive_main_file_detection has been renamed to g:vimtex_disable_directory_scan_main_file_detection.
Performance evaluation for the work on PR #2558 gave evidence that disabling the directory scan method for the main file detection due to performance concerns appears to be unjustified. Remove this option. Note that recently g:vimtex_disable_recursive_main_file_detection has been renamed to g:vimtex_disable_directory_scan_main_file_detection.
As a result of #2465, I added a restriction to the compiler module where it won't compile the document unless the main file was verified to be a valid main file. This helped to ensure robustness, at the cost of breaking the work flow for several users.
This PR introduces a minor change to how VimTeX checks if a file is a main file that should resolve most of the issues reported after the more restrictive compiler module. However, I won't merge it immediately, because it may come at a cost of reduced init performance. That is, slower load time for VimTeX. This should only affect large projects, and only if they consist of many/multiple files. But still, it may be a relevant blocker.
Therefore, before I merge this, I would be very happy if users who have large LaTeX documents with multiple files (e.g. books, large theses, etc) could test if this branch introduce a noticable reduced startup time. Please also report if you do not see a reduced performance.
Related issues:
People who may have an interest in this issue: @Rmano, @fauust, @Yorkwn, @shuber2, @krishnakumarg1984, @clason