Extremely high memory usage when opening a large project

[koby08]

Hi,

I've been trying to use the language server in a quite big private library (~10MB of code), and the language server indeed indexed all the R files. However, its memory consumption reached ~12GB.

Is there a way I could minimize the footprint?

Thanks

[renkun-ken]

languageserver currently will parse all R files in R folder in the workspace if it contains NAMESPACE file, indicating that it is an R package. Is it your case?

[Ohmegas]

Yes, it's a quite big library. In RStudio I didn't see any issues or high memory usage, but I really want to give vscode try

[renkun-ken]

Would you like to give us some simple numbers to describe how large is the package? Number of files in R, average file size, etc? Looks like we should give up parsing all files in R if we find the folder contains too many files, or there might be smarter behavior so that we only parse a subset of files.

[Ohmegas]

Sure - few hundreds of files, ~10mb of code, pretty much evenly distributed between them..

Hmm, it is a relatively big project I assume, but I dont really understand how come loading 10mb of code becomes 12gb of RAM? Any idea? RStudio for comparison seems to work well in this project, and takes ~100mb of RAM

[Ohmegas]

I wrote a small patch that instead of relying on the parsed files that are loaded in workspace$load_all - it extracts the srcref of the function that was received in a definition request, and then parses the relevant file adhoc, and then tries to resolve the definition.

It works pretty well, but still using a hacky implementation.

What do you think about such a feature for the server? (Of course with a proper implementation)

[koby08]

@renkun-ken Any feedback on this?

[Tal500]

I'm working at the same company this issue happens. I've opened a detailed issue and suggestion in #618