How can I use these, getOrCreateDocument and initializeWorkspace in CLI?

[negahama]

Hello everyone
I made my CLI, It works well, but It has a big serious problem. It's too slow.
Most of the time is spent on langiumDocuments.getOrCreateDocument(). 99% of the total working time is spent here really.

And I made My CLI base on #861, so every time my CLI transpile the files, It calls workspace.WorkspaceManager.initializeWorkspace()

I'm wondering why getOrCreateDocument() is taking so much time, but I'm more curious about how to solve it than why.
How could I not call initializeWorkspace() every time for performance?
Thanks.

[msujew]

Hey @negahama,

Can you share some of your code with us? Without more info, it's very hard to figure out how to fix your performance issue. Have you tried profiling your CLI yet? There are often times performance bottlenecks im the scoping and validation implementations.

[negahama]

Hi @msujew

My project is Here. You can try to it

npm install
cd scala-script-cli
npm run all

then my cli program will generate files to scala-script-cli/samples/generated/ from the sample files in scala-script-cli/samples/
You could see the result message in terminal in a minute
As I told you, the most of time you wait is spent getOrCreateDocument()
Thank you for help

[msujew]

Looking at this through a profiler, it seems like the parser is taking a very long time on your files - mainly with attempting to perform the lookahead. This indicates very long common prefixes in the grammar:

Looking at your grammar, it seems like your language features a lot of ambiguity in the first few tokens. For example ForOf and ForTo can be joined into a single rule, and everything that starts with (Annotation)? (export?='export')? (private?='private')? (static?='static')? can be moved into a single prefixed rule that then fans out into different branches. Also, the lookahead creates a warning in your code saying that Block is completely ambiguous:

Ambiguous Alternatives Detected: <0, 1> in <OR1> inside <Block​> Rule,
<{, }, }> may appears as a prefix path in all these alternatives.
See: https://chevrotain.io/docs/guide/resolving_grammar_errors.html#AMBIGUOUS_ALTERNATIVES
For Further details.

[negahama]

@msujew Thank you so much...
I have something I want to make clear.
I thought getOrCreateDocument() have nothing to do with parsing... Is it right?
Are Lexing and parsing processed in DocumentFactory.build?

[msujew]

Creating a document automatically parses it's content. There is no document without content.

[negahama]

I see... that is why getOrCreateDocument takes so long time... thank you!!

And one more...
Whenever I want to build, Do I have to call initializeWorkspace in CLI?
Is there better way to do that?

[msujew]

I would recommend to use the DocumentBuilder.build API to build documents. The WorkspaceManager is using that API internally as well.