How do I compile tree-sitter parsers? - gcc

Question

I'm trying to compile the typescript, tsx, and python parsers for use with neovim, but when I run gcc -o parser.so -shared src/parser.c -Os -I./src as instructed by the tree-sitter tracking issue I get:
Undefined symbols for architecture x86_64:
"_tree_sitter_tsx_external_scanner_create", referenced from:
_tree_sitter_tsx.language in parser-eb4e9b.o
"_tree_sitter_tsx_external_scanner_deserialize", referenced from:
_tree_sitter_tsx.language in parser-eb4e9b.o
"_tree_sitter_tsx_external_scanner_destroy", referenced from:
_tree_sitter_tsx.language in parser-eb4e9b.o
"_tree_sitter_tsx_external_scanner_scan", referenced from:
_tree_sitter_tsx.language in parser-eb4e9b.o
"_tree_sitter_tsx_external_scanner_serialize", referenced from:
_tree_sitter_tsx.language in parser-eb4e9b.o
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
I've run the npm build steps, and tried the make file in build, but I'm not able to get the final .so files.

Answer 1

the error undefined symbol: tree_sitter_parsername_external_scanner_create also appears, when there is a mismatch between parser name and function names
// grammar.js
name: 'parsername',
// scanner.cc
void *tree_sitter_oldparsername_external_scanner_create() {
return new Scanner();
}
quickfix: replace name in scanner source
sed -i -E 's|(tree_sitter)_oldparsername_|\1_parsername_|g' src/scanner.*

Answer 2

Some grammars have external scanners - hand written C/C++ source files that are by convention called scanner.c or scanner.cc, which need to be compiled along with parser.c.