v/vlib/v
Delyan Angelov ccb5f1d563
tests: support `// vtest vflags: -prod` in `coutput_test.v`.
Use it, to add a `-prod` test for `$embed_file()`,
to prevent future codegen regressions.
2021-10-05 15:44:12 +03:00
..
ast v.ast: fix fmt of 'bytestr' to '&byte' (#12014) 2021-09-29 13:53:46 +03:00
builder v.builder,v.gen: time Builder.front_stages.parse_files, Builder.resolve_deps, Gen.sort_structs 2021-09-30 20:49:19 +03:00
callgraph v: support -show-depgraph in addition to -show-callgraph 2021-07-28 16:42:11 +03:00
cflag
checker v.checker: fix return type checking being skipped for mutable method receivers (#12043) 2021-10-02 23:26:46 +03:00
depgraph v: support -show-depgraph in addition to -show-callgraph 2021-07-28 16:42:11 +03:00
doc all: more chdir/chmod fixes 2021-08-28 12:52:48 +03:00
dotgraph v: support -show-depgraph in addition to -show-callgraph 2021-07-28 16:42:11 +03:00
embed_file
errors v.builder: implement a `-check` mode, that runs only the parser + the checker, without codegen (#11414) 2021-09-07 07:17:53 +03:00
eval
fmt fmt: fix import with symbols (fix #12065) (#12069) 2021-10-05 11:44:48 +03:00
gen tests: support `// vtest vflags: -prod` in `coutput_test.v`. 2021-10-05 15:44:12 +03:00
live tests: support `// vtest retry: 4` for marking flaky tests 2021-09-18 11:48:32 +03:00
markused v.markused: remove `hash.init()` function (map fn usages are detected more precisely) 2021-09-23 12:24:02 +03:00
parser cgen: fix threads array wait without go calls (fix #12009) (#12050) 2021-10-03 18:24:44 +03:00
pkgconfig fmt: fix unnecessary line break in array init (fix #11448) (#11562) 2021-09-21 16:20:09 +03:00
pref v.pref: disable the parallel cgen when `-g` is given for now (the line numbers/paths are wrong) 2021-10-05 15:08:28 +03:00
preludes builtin: fix leak in rune.str(), fix leaks in most assert x == y statements in tests (#11091) 2021-08-13 18:37:34 +03:00
preludes_js v.gen.js: refactor code, fix alias codegen, `-stats` now again works with tests on the JS backend (#11512) 2021-09-16 14:00:15 +03:00
scanner cgen, fmt, scanner: fix and use nested lambda in scanner (#11967) 2021-09-24 18:21:22 +03:00
tests tests: rename `as_cast_literal.v` to `as_cast_literal_test.v` 2021-10-05 11:57:57 +03:00
token all: support `mut volatile x := 123` declarations (#11940) 2021-09-22 13:06:30 +03:00
transformer v: add constant folding AST transformation (#11085) 2021-08-06 23:19:18 +03:00
util 0.2.4 2021-08-30 22:25:14 +03:00
vcache os: make chmod() return error 2021-08-28 09:37:24 +03:00
vet
vmod v.vmod: return errors instead of panic-ing 2021-08-16 16:54:41 +03:00
README.md
TEMPLATES.md tools: improve `v check-md` by checking for broken TOC headline links (#10417) 2021-06-14 13:12:02 +03:00
compiler_errors_test.v fmt: fix unnecessary line break in array init (fix #11448) (#11562) 2021-09-21 16:20:09 +03:00

README.md

Compiler pipeline

A simple high level explanation how the compiler pipeline (parser -> checker -> generator) works.

Reading files

Getting builtin files

To load all builtin files, a preference Preferences.lookup_path for the path where to look for exists. See Builder.get_builtin_files as example. If the file is a .vsh file and the backend is C, vlib/os will also be loaded as builtin.

Getting project files

Either there is a specific file: my_file.v or a directory containing V files. In the last case it scans that directory for all files. See Builder.v_files_from_dir as the helper method. This list of files needs to be filtered so that only *.v files exist.

Skips the following file types:

  • *_test.v
  • either *.c.v or *.c.js depending on the backend
  • all files that doesn't end with .v
  • Files that are not defined in Preferences.compile_defines or Preferences.compile_defines_all if any file is defined.

Parsing files

To parse something a new template is created as the first step:

import v.ast

table := ast.new_table()

a new preference is created:

import v.pref

pref := &pref.Preferences{}

and a new scope is created:

import v.ast

scope := ast.Scope{
	parent: 0
}

after that, you can parse your files.

Parse text

If you want to parse only text which isn't saved on the disk you can use this function.

import v.parser

code := ''
// table, pref and scope needs to be passed as reference
parsed_file := parser.parse_text(code, table, .parse_comments, &pref, &scope)

Parse a single file

For parsing files on disk, a path needs to be provided. The paths are collected one step earlier.

import v.parser

path := ''
// table, pref and scope needs to be passed as reference
parsed_file := parser.parse_file(path, table, .parse_comments, &pref, &scope)

Parse a set of files

If you have a batch of paths available which should be parsed, there is also a function which does all the work.

import v.parser

paths := ['']
// table, pref and scope needs to be passed as reference
parsed_files := parser.parse_files(paths, table, &pref, &scope)

Parse imports

A file often contains imports. These imports might need to be parsed as well. The builder contains a method which does this: Builder.parse_imports.

If the module which is imported isn't parsed already, you have to collect it relatively from the main file. For this the ast.File contains a list of imports. Those imports needs to be found on disk. . is just replaced with seperators in the relative location of the main file. Then all files from that directory are collected and parsed again like the previous steps explained.

Checking AST

A new checker is created:

import v.checker

mut checker := checker.new_checker(table, &pref)

After checking your files in checker.errors and checker.warnings you can see the results.

Check ast.File

checker.check(parsed_file)

Check a list of ast.File

checker.check_files(parsed_files)

Generate target from AST

Generating C code works just as this:

import v.gen.c

res := c.gen(parsed_files, table, &pref)