tree-sitter: Main thread panicks with 'no entry found for key'
I have created a grammar which causes tree-sitter generate to fail with the following error:
$ /home/cu3po42/development/tree-sitter-minerva/node_modules/.bin/tree-sitter generate
thread 'main' panicked at 'no entry found for key', cli/src/generate/render.rs:1049:21
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
error Command failed with exit code 101.
The verbose backtrace is as follows:
$ /home/cu3po42/development/tree-sitter-minerva/node_modules/.bin/tree-sitter generate
thread 'main' panicked at 'no entry found for key', cli/src/generate/render.rs:1049:21
stack backtrace:
0: 0x564c8314c495 - backtrace::backtrace::libunwind::trace::h14d338b30b3ea0a7
at /cargo/registry/src/github.com-1ecc6299db9ec823/backtrace-0.3.46/src/backtrace/libunwind.rs:86
1: 0x564c8314c495 - backtrace::backtrace::trace_unsynchronized::h73ea91d74a3fd67f
at /cargo/registry/src/github.com-1ecc6299db9ec823/backtrace-0.3.46/src/backtrace/mod.rs:66
2: 0x564c8314c495 - std::sys_common::backtrace::_print_fmt::hd42948c952866e12
at src/libstd/sys_common/backtrace.rs:78
3: 0x564c8314c495 - <std::sys_common::backtrace::_print::DisplayBacktrace as core::fmt::Display>::fmt::ha8f928866ff7571e
at src/libstd/sys_common/backtrace.rs:59
4: 0x564c83177f0c - core::fmt::write::he0c1e5f7426d2718
at src/libcore/fmt/mod.rs:1076
5: 0x564c83144622 - std::io::Write::write_fmt::hf3afc6cfd57d0033
at src/libstd/io/mod.rs:1537
6: 0x564c8314ecd0 - std::sys_common::backtrace::_print::hfc0110703f3696fd
at src/libstd/sys_common/backtrace.rs:62
7: 0x564c8314ecd0 - std::sys_common::backtrace::print::h3f77c6990ddfaa22
at src/libstd/sys_common/backtrace.rs:49
8: 0x564c8314ecd0 - std::panicking::default_hook::{{closure}}::heae49580a8d62d75
at src/libstd/panicking.rs:198
9: 0x564c8314ea1c - std::panicking::default_hook::hecc34e3f729e213c
at src/libstd/panicking.rs:217
10: 0x564c8314f313 - std::panicking::rust_panic_with_hook::he82f5d0644692441
at src/libstd/panicking.rs:526
11: 0x564c8314ef0b - rust_begin_unwind
at src/libstd/panicking.rs:437
12: 0x564c831768a1 - core::panicking::panic_fmt::h09c929f06bb87c98
at src/libcore/panicking.rs:85
13: 0x564c83176633 - core::option::expect_failed::hb49e05615ae76cc6
at src/libcore/option.rs:1261
14: 0x564c82f1e544 - tree_sitter_cli::generate::render::Generator::add_parse_table::h69cfb8cfa7f865a7
15: 0x564c82f124f1 - tree_sitter_cli::generate::render::Generator::generate::h1a94bdd55f338bfc
16: 0x564c82f25e4a - tree_sitter_cli::generate::render::render_c_code::h100f4f082d98533d
17: 0x564c82ed4873 - tree_sitter_cli::generate::generate_parser_for_grammar_with_opts::h4a8f7bf507ee7555
18: 0x564c82ed28ef - tree_sitter_cli::generate::generate_parser_in_directory::hfb424fe8b81fc43a
19: 0x564c82e83fe6 - tree_sitter::run::h6a21535c52bdfbe2
20: 0x564c82e811f2 - tree_sitter::main::hd24ee9c5bfe24a1c
21: 0x564c82e882c3 - std::rt::lang_start::{{closure}}::hffb19e98173c15bc
22: 0x564c8314f6e3 - std::rt::lang_start_internal::{{closure}}::h5d3ea623498f5f43
at src/libstd/rt.rs:52
23: 0x564c8314f6e3 - std::panicking::try::do_call::hac65e71be769a440
at src/libstd/panicking.rs:348
24: 0x564c8314f6e3 - std::panicking::try::hd4706e264bcf6712
at src/libstd/panicking.rs:325
25: 0x564c8314f6e3 - std::panic::catch_unwind::h948a0fb4a8b3ee82
at src/libstd/panic.rs:394
26: 0x564c8314f6e3 - std::rt::lang_start_internal::h72cc068ed2d0ac53
at src/libstd/rt.rs:51
27: 0x564c82e882b2 - main
28: 0x7fe65fda91e3 - __libc_start_main
29: 0x564c82e7f839 - _start
30: 0x0 - <unknown>
error Command failed with exit code 101.
I tried to create a minimal grammar that exhibits this error and arrived at the following:
module.exports = grammar({
name: 'NAME',
rules: {
source_file: $ => "foo",
comment: $ => $.bar,
bar: $ => seq(
"bar",
),
},
extras: $ => [$.comment],
});
Performing any of the following will “fix” the problem:
- Removing
extras - Setting
source_fileto$.comment - Replacing
seq("bar")by"bar" - Manually inlining the reference to
bar
I therefore assume that while processing the rules, the process gets to neither comment nor bar, since they are not referenced from source_file, and then when checking extras, the rule map isn’t populated in the way it should be.
However, upon reviewing the code I found this, which seems to suggest that the extras rules are used as an entry point, so I’m not entirely sure what’s going on.
For now I was able to circumvent this issue by allowing comments to be used directly on the top level.
About this issue
- Original URL
- State: closed
- Created 6 years ago
- Reactions: 6
- Comments: 34 (20 by maintainers)
Commits related to this issue
- trying to parse comments (blocked by https://github.com/tree-sitter/tree-sitter/issues/768 ) — committed to rems-project/tree-sitter-core by vzaliva 2 years ago
If you want to view the issue tracker as a liability then whatever. I won’t be reporting further issues to this project.
All right, I will stop being so hot-headed about it. We are all volunteers here. I will go back through the issues I’ve reported and figure out whether they’re still valid.
@vzaliva apologies if you already did but did you try removing
$.commentand putting$.line_commentand$.block_commentinextrasdirectly? This is what solved my issue (see above).If that doesn’t work you can bisect the problem. Remove some parts from your grammar and patch up the rest so that it should generate. If you get the “no entry found for key” error remove more, if you don’t get the error add some back, if you get a different error fix it and retry.
Here it is: https://github.com/rems-project/tree-sitter-core/blob/main/grammar.js
The problem started when I did more elaborate comments parsing (starting from the
commentrule).yes, GH auto closed it
I can confirm I get this in 0.20.8, 0.20.9, and 0.21.0-pre-release-1.
The cause seems to be a
choicerule inextras. I had:and changing to
resolved the issue. Furthermore, the symbol in question which didn’t have a symbol_id was
line_comment.Interestingly, tree-sitter-rust seems to have a workaround for this exact issue: it has an unused
commentrule, andextrascontains$.line_commentand$.block_comment.Only the ones that are super-old and I suspect it might have already been fixed. Do you think this approach is incorrect somehow?
@alanz I’ve encountered issues with an accidental duplicate rule name before. Would be nice to have a warning for that.