Improve performance in large files #1507
Merged
+49
−9
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This shows up *heavily* in profiles. What really needs to happen is necessary info is precomputed and parsed once after we recieve edits for files and then reused everywhere *until the next edit*. That’s a larger refactor so this is a stop gap measure until then.
This will help perf in two cases: 1. Repeated compilation of the same classes across requests 2. Compilation *of the same* class mulitple times in a single call We store this data on the design system itself so that it gets invalidated and garbage collected when the design system is recreated on CSS file changes
thecrypticace
force-pushed
the
fix/issue-1503
branch
from
e23c73a to
b75417f
Compare
thecrypticace
changed the title
timsexperiments
pushed a commit
to timsexperiments/tailwindcss-intellisense
that referenced
this pull request
Dec 4, 2025
* Bump bundled Tailwind CSS version * Improve performance in large files (tailwindlabs#1507) There are two performance improvements here that both come down to caching: - We call `indexToPosition` *a lot*. Every time this is called it takes the string passed in and re-computes the indices every time. There's no need for this so we have a small, module-level LRU cache to store the data. This should eventually be replaced with parsing and storing data up front when files are edited instead of just doing this whenever we receive a request. That would negate the need for a module-level cache. - We call `compile()` and `candidatesToCss` *a lot* as well. This is expected but 90% of the time the classes passed in have already been seen. We take the generated CSS and convert to a PostCSS node because a lot of the implementation uses that internally. There's no need to go through the whole parse -> serialize -> re-parse routine like we are. I've opted to store the compiled PostCSS roots on the `designSystem.storage` object (which will be created when it doesn't exist on old versions) and we'll return clones of those objects every time (since they may get mutated). With this there's only one large-ish block in a left-heavy profile: `findClassListsInHtmlRange`. That is mostly related to the class attribute lexer. Fixing it will require us to scan document and store info about them when they change. We shouldn't need to run the class attribute lexer if the document hasn't changed. Fixing this will require some more work though. Fixes tailwindlabs#1503 --------- Co-authored-by: Jordan Pittman <jordan@cryptica.me> Co-authored-by: GitHub Action <action@github.com>
timsexperiments
pushed a commit
to timsexperiments/tailwindcss-intellisense
that referenced
this pull request
Dec 5, 2025
* Bump bundled Tailwind CSS version * Improve performance in large files (tailwindlabs#1507) There are two performance improvements here that both come down to caching: - We call `indexToPosition` *a lot*. Every time this is called it takes the string passed in and re-computes the indices every time. There's no need for this so we have a small, module-level LRU cache to store the data. This should eventually be replaced with parsing and storing data up front when files are edited instead of just doing this whenever we receive a request. That would negate the need for a module-level cache. - We call `compile()` and `candidatesToCss` *a lot* as well. This is expected but 90% of the time the classes passed in have already been seen. We take the generated CSS and convert to a PostCSS node because a lot of the implementation uses that internally. There's no need to go through the whole parse -> serialize -> re-parse routine like we are. I've opted to store the compiled PostCSS roots on the `designSystem.storage` object (which will be created when it doesn't exist on old versions) and we'll return clones of those objects every time (since they may get mutated). With this there's only one large-ish block in a left-heavy profile: `findClassListsInHtmlRange`. That is mostly related to the class attribute lexer. Fixing it will require us to scan document and store info about them when they change. We shouldn't need to run the class attribute lexer if the document hasn't changed. Fixing this will require some more work though. Fixes tailwindlabs#1503 --------- Co-authored-by: Jordan Pittman <jordan@cryptica.me> Co-authored-by: GitHub Action <action@github.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There are two performance improvements here that both come down to caching:
indexToPositiona lot. Every time this is called it takes the string passed in and re-computes the indices every time. There's no need for this so we have a small, module-level LRU cache to store the data.This should eventually be replaced with parsing and storing data up front when files are edited instead of just doing this whenever we receive a request. That would negate the need for a module-level cache.
compile()andcandidatesToCssa lot as well. This is expected but 90% of the time the classes passed in have already been seen. We take the generated CSS and convert to a PostCSS node because a lot of the implementation uses that internally. There's no need to go through the whole parse -> serialize -> re-parse routine like we are.I've opted to store the compiled PostCSS roots on the
designSystem.storageobject (which will be created when it doesn't exist on old versions) and we'll return clones of those objects every time (since they may get mutated).With this there's only one large-ish block in a left-heavy profile:
findClassListsInHtmlRange. That is mostly related to the class attribute lexer. Fixing it will require us to scan document and store info about them when they change. We shouldn't need to run the class attribute lexer if the document hasn't changed. Fixing this will require some more work though.Fixes #1503