Optimizing Performance

Working with PSI efficiently

Avoid PsiElement methods which are expensive with deep trees

getText() traverses the whole tree under the given element and concatenates strings, consider textMatches() instead.

getTextRange(), getContainingFile(), getProject() traverse the tree up to the file, which can be long in very nested trees. If you only need PSI element length, use getTextLength().

File and project often can be computed once per some analysis and then stored in fields or passed via parameters.

Additionally, getText(), getNode(), getTextRange(), etc., all need AST, which can be quite an expensive operation. See below.

Avoid loading too many parsed trees or documents into memory at the same time

Ideally, only AST nodes from files open in the editor should be present in the memory. Everything else, even if it’s needed for resolve/highlighting purposes, can be accessed via PSI interfaces, but its implementations should use stubs underneath, which are less CPU- and memory-expensive.

If stubs don’t suit your case well (e.g., the information you need is large and/or very rarely needed, or you’re developing a plugin for a language whose PSI you don’t control), you can create a custom index or gist.

You can use AstLoadingFilter in production and PsiManagerEx.setAssertOnFileLoadingFilter() in tests to ensure you’re not loading AST accidentally.

The same applies to documents: only the ones opened in editors should be loaded. Usually, you shouldn’t need document contents (as most information can be retrieved from PSI). If you nevertheless need documents, consider saving the information you need to provide in a custom index or gist to get it more cheaply later. If you still need documents, then at least ensure you load them one by one and don’t hold them on strong references to let GC free the memory as quickly as possible.

Cache results of heavy computations

These include PsiElement.getReference(s), PsiReference.resolve() (and multiResolve() and other equivalents), expression types, type inference results, control flow graphs, etc.

Usually, CachedValue works well.

If the information you cache depends only on a subtree of the current PSI element (and nothing else: no resolve results or other files), you can cache it in a field in that PsiElement and drop the cache in an override of ASTDelegatePsiElement.subtreeChanged().

Improving indexing performance

Avoid using AST

Use lexer information instead of parsed trees if possible.

If impossible, use light AST which doesn’t create memory-hungry AST nodes inside, so traversing it might be faster. Make sure to traverse only the nodes you need to.

For stub index, implement LightStubBuilder. For other indices, you can get the light AST manually via ((PsiDependentFileContent) fileContent).getLighterAST().

Consider prebuilt stubs

If your language has a massive standard library, which is mostly the same for all users, you can avoid stub-indexing it in each installation by providing prebuilt stubs with your distribution. See PrebuiltStubsProvider extension.

Avoiding UI freezes

Don’t perform long operations in UI thread

In particular, don’t traverse VFS, parse PSI, resolve references or query FileBasedIndex.

There are cases when the platform itself invokes such expensive code (e.g., resolve in AnAction.update()). We’re trying to eliminate them. Meanwhile, you can try to speed up what you can in your language, it’ll be beneficial anyway, as it’ll also improve background highlighting performance.

WriteActions currently have to happen on UI thread, so to speed them up, you can try moving as much as possible out of write action into a preparation step which can be then invoked in background (e.g., using ReadAction.nonBlocking()).

Don’t do anything expensive in event listeners. Ideally, you should only clear some caches. You can also schedule background processing of events, but be prepared that some new events might be delivered before your background processing starts, and thus the world might have changed by that moment or even in the middle of background processing. Consider using MergingUpdateQueue and ReadAction.nonBlocking() to mitigate these issues.

Massive batches of VFS events can be pre-processed in background, see AsyncFileListener (2019.2 or later).

Don’t block EDT by long non-cancellable ReadActions in background threads

See General Threading Rules, especially its section on Read Action Cancellability.