3 unstable releases
0.2.1 | Apr 22, 2023 |
---|---|
0.2.0 | Apr 22, 2023 |
0.1.0 | Apr 21, 2023 |
#270 in #tree-sitter
17KB
309 lines
Code-shape
Code-shape is a tool that uses Tree-sitter to extract a shape of code definitions from a source code file. The tool uses the same language parsers that are installed for Tree-sitter CLI.
Installation
To install the code-shape
CLI it's possible to use Rust's Cargo package manager:
cargo install code-shape
Usage
To start using the tool it's needed to do some preparation.
Prerequsites
- Install Tree-sitter CLI.
- Run
tree-sitter init-config
that creates a config file like~/.config/tree-sitter/config.json
in Tree-sitter's config dir. - Create a directory where installed parsers would be located and add it in
"parser-directories"
list in Tree-sitter's config file. - Clone Tree-sitter parsers for required languages to the parsers directory.
Define extraction query
To make it possible to extract a shape of definitions from some source code file for some language, it's needed to define a query. To define a new query create a file in a Code-shape's languages config dir ~/.config/code-shape/languages/
with an .scm
suffix like ~/.config/code-shape/languages/c.scm
and put there a set of Tree-sitter query patterns like:
; C language function declarations
(declaration
[
(function_declarator
declarator: (identifier) @fn.declaration.name
)
(_
(function_declarator
declarator: (identifier) @fn.declaration.name
)
)
]
)
; C language function pointer declarations
(declaration
[
(init_declarator
(function_declarator
(_ (_ declarator: (identifier) @fn.pointer.declaration.name))
)
)
(init_declarator
(_
(function_declarator
(_ (_ declarator: (identifier) @fn.pointer.declaration.name))
)
)
)
]
)
; C language function definitions
(function_definition
[
(function_declarator
declarator: (identifier) @fn.name
)
(_
(function_declarator
declarator: (identifier) @fn.name
)
)
]
body: (_) @fn.scope
)
It's needed to define captures with special names:
<type>.name
is a capture where thetype
may be, e.g.,fn
,class
or anything else to match a code entity name.<type>.scope
is a special capture that allows for the tool to capture a context of entities and usually are tokens that defines a body of the the entity, e.g., a function body.
Examples of the tool output:
# code-shape --scope source.c tree-sitter/lib/src/alloc.c
fn ts_malloc_default
fn ts_calloc_default
fn ts_realloc_default
fn.pointer.declaration ts_current_malloc
fn.pointer.declaration ts_current_calloc
fn.pointer.declaration ts_current_realloc
fn.pointer.declaration ts_current_free
fn ts_set_allocator
# code-shape examples/foo.c
fn.declaration ts_malloc_default
fn.declaration ts_calloc_default
fn.declaration ts_realloc_default
fn.pointer.declaration ts_current_malloc
fn.pointer.declaration ts_current_calloc
fn.pointer.declaration ts_current_realloc
fn.pointer.declaration ts_current_free
fn foo
fn bar
# code-shape examples/foo.py
class Foo
def foo
def bar
def inner
def one
def two
def wrap
class Baz
class Bar
class Foo
def func1
def func2
def three
def four
Embedded shape queries
For now the tool has builtin shape queries for the following language parsers:
Dependencies
~10–21MB
~299K SLoC