No highlighting on large files

[Jomy10]

On large files highlighting stops working. I'm not sure where the limitation is enforced (zed, tree sitter or this extension?)

Example:

[harmohitsandhu]

+1
I was just working on a large XML file and Zed stopped highlighting syntax. Highlighting for large XML files works fine on VSCodium though.

[sweetppro]

@harmohitsandhu @Jomy10 does v0.0.9 fix this issue

[harmohitsandhu]

@sweetppro I tested with v0.0.9, but unfortunately the issue still persists.

[harmohitsandhu]

@sweetppro Okay, it looks like most of the XML content is highlighted now. I work with Blogger templates, which are XML-based and include embedded CSS, JavaScript, and structured markup. In VSCodium, the entire file including CSS, JavaScript, and the structured markup is highlighted. However, in Zed, only the structured markup is highlighted, while the CSS and JavaScript sections appear unhighlighted.

[sweetppro]

@harmohitsandhu nice, so a slight improvement 😄

[jwhitaker-gridcog]

This still seems to be an issue, I'm hitting it with Tableau workbooks (which are.. pretty big, order of MiB).

Here's one that is reproducing the blank highlighting: Student Loans in the USA.twb.xml

Picked at random from tableau public - https://public.tableau.com/app/profile/ingibjorg.jakobsdottir/viz/StudentLoansintheUSA/CapMap . Most of the entries on public are too small to reproduce this (and many have downloading the raw workbook disabled), but this one was big enough and open. I'm not sure if the problem is triggered by raw size, or complexity of document structure, but either way this one is about 600KiB uncompressed. Internally I'm working on one that's about 2MiB.

[wezm]

I can reproduce this. Zed reports:

2026-01-05T10:34:49+10:00 ERROR [language::syntax_map] error parsing text: failed to parse

Attempting to parse the xml file with the xml.wasm file results in the same error. With some debug output enabled in my test project I see:

     Running `target/debug/zed-ts-wasm ../../Downloads/Student.Loans.in.the.USA.twb.xml`
trap in wasm module: error while executing at wasm backtrace:
    0:  0x649 - stdlib.wasm!memcpy
    1: 0x4900 - <unknown>!tree_sitter_xml_external_scanner_serialize

Caused by:
    wasm trap: out of bounds memory access

thread 'main' (926216) panicked at src/main.rs:23:50:
unable to parse

This might be an issue with scanner.c in tree-sitter-xml, although it does not happen when using the tree-sitter C API, only the wasm module.

I spent a bunch of time trying to work out why/how the memcpy fails. I think I narrowed it down to the memcpy inside the loop in scanner.c, but not much more than that.

Finally I tried updating the tree-sitter crate from the version used in the currently the released versions of Zed to the latest one, and the parser started working in my test project. I haven't worked out what was fixed/changed in tree-sitter. There is a commit on Zed main to bump tree-sitter to the latest version. zed-industries/zed#44963 This commit does not appear to have made it to a Zed release yet, but I built it form source and can confirm that it does fix highlighting for me:

[Jomy10]

wasm trap: out of bounds memory access means wasm doesn't have enough memory if I remember correctly.

Not sure if Zed allows wasm plugins to grow the allocated memory or not.

I'd guess the latest xml tree-sitter just requires less memory to run?

[wezm]

wasm trap: out of bounds memory access means wasm doesn't have enough memory if I remember correctly.

Not sure if Zed allows wasm plugins to grow the allocated memory or not.

I'd guess the latest xml tree-sitter just requires less memory to run?

Yes that's possible. However, the xml plugin checks the write is within the limit before calling memcpy, so I would not expect the trap to happen.

https://github.com/tree-sitter-grammars/tree-sitter-xml/blob/863dbc381f44f6c136a399e684383b977bb2beaa/xml/src/scanner.c#L218-L224

[tlrib]

Having the same issue on a 10mb xml file

[koenvb]

@wezm would there be any alternative for this? Would love to have xml just properly working, any help or so I could provide to make this working or any alternative way?

[wezm]

would there be any alternative for this?

Not really. The structure of this plugin is how pretty much all Zed language extensions work.

any help or so I could provide to make this working or any alternative way?

The main thing needed is trying to reproduce the issue and track down the source. Also making problematic sample documents available for testing would be useful.