path: root/internal/confyg/read.go

// Copyright 2018 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// Module file parser.
// This is a simplified copy of Google's buildifier parser.

package confyg

import (
	"bytes"
	"fmt"
	"os"
	"strings"
	"unicode"
	"unicode/utf8"
)

// A Position describes the position between two bytes of input.
type Position struct {
	Line     int // line in input (starting at 1)
	LineRune int // rune in line (starting at 1)
	Byte     int // byte in input (starting at 0)
}

// add returns the position at the end of s, assuming it starts at p.
func (p Position) add(s string) Position {
	p.Byte += len(s)
	if n := strings.Count(s, "\n"); n > 0 {
		p.Line += n
		s = s[strings.LastIndex(s, "\n")+1:]
		p.LineRune = 1
	}
	p.LineRune += utf8.RuneCountInString(s)
	return p
}

// An Expr represents an input element.
type Expr interface {
	// Span returns the start and end position of the expression,
	// excluding leading or trailing comments.
	Span() (start, end Position)

	// Comment returns the comments attached to the expression.
	// This method would normally be named 'Comments' but that
	// would interfere with embedding a type of the same name.
	Comment() *Comments
}

// A Comment represents a single // comment.
type Comment struct {
	Start  Position
	Token  string // without trailing newline
	Suffix bool   // an end of line (not whole line) comment
}

// Comments collects the comments associated with an expression.
type Comments struct {
	Before []Comment // whole-line comments before this expression
	Suffix []Comment // end-of-line comments after this expression

	// For top-level expressions only, After lists whole-line
	// comments following the expression.
	After []Comment
}

// Comment returns the receiver. This isn't useful by itself, but
// a Comments struct is embedded into all the expression
// implementation types, and this gives each of those a Comment
// method to satisfy the Expr interface.
func (c *Comments) Comment() *Comments {
	return c
}

// A FileSyntax represents an entire go.mod file.
type FileSyntax struct {
	Name string // file path
	Comments
	Stmt []Expr
}

func (x *FileSyntax) Span() (start, end Position) {
	if len(x.Stmt) == 0 {
		return
	}
	start, _ = x.Stmt[0].Span()
	_, end = x.Stmt[len(x.Stmt)-1].Span()
	return start, end
}

// A CommentBlock represents a top-level block of comments separate
// from any rule.
type CommentBlock struct {
	Comments
	Start Position
}

func (x *CommentBlock) Span() (start, end Position) {
	return x.Start, x.Start
}

// A Line is a single line of tokens.
type Line struct {
	Comments
	Start Position
	Token []string
	End   Position
}

func (x *Line) Span() (start, end Position) {
	return x.Start, x.End
}

// A LineBlock is a factored block of lines, like
//
//	require (
//		"x"
//		"y"
//	)
//
type LineBlock struct {
	Comments
	Start  Position
	LParen LParen
	Token  []string
	Line   []*Line
	RParen RParen
}

func (x *LineBlock) Span() (start, end Position) {
	return x.Start, x.RParen.Pos.add(")")
}

// An LParen represents the beginning of a parenthesized line block.
// It is a place to store suffix comments.
type LParen struct {
	Comments
	Pos Position
}

func (x *LParen) Span() (start, end Position) {
	return x.Pos, x.Pos.add(")")
}

// An RParen represents the end of a parenthesized line block.
// It is a place to store whole-line (before) comments.
type RParen struct {
	Comments
	Pos Position
}

func (x *RParen) Span() (start, end Position) {
	return x.Pos, x.Pos.add(")")
}

// An input represents a single input file being parsed.
type input struct {
	// Lexing state.
	filename  string    // name of input file, for errors
	complete  []byte    // entire input
	remaining []byte    // remaining input
	token     []byte    // token being scanned
	lastToken string    // most recently returned token, for error messages
	pos       Position  // current input position
	comments  []Comment // accumulated comments
	endRule   int       // position of end of current rule

	// Parser state.
	file       *FileSyntax // returned top-level syntax tree
	parseError error       // error encountered during parsing

	// Comment assignment state.
	pre  []Expr // all expressions, in preorder traversal
	post []Expr // all expressions, in postorder traversal
}

func newInput(filename string, data []byte) *input {
	return &input{
		filename:  filename,
		complete:  data,
		remaining: data,
		pos:       Position{Line: 1, LineRune: 1, Byte: 0},
	}
}

// parse parses the input file.
func parse(file string, data []byte) (f *FileSyntax, err error) {
	in := newInput(file, data)
	// The parser panics for both routine errors like syntax errors
	// and for programmer bugs like array index errors.
	// Turn both into error returns. Catching bug panics is
	// especially important when processing many files.
	defer func() {
		if e := recover(); e != nil {
			if e == in.parseError {
				err = in.parseError
			} else {
				err = fmt.Errorf("%s:%d:%d: internal error: %v", in.filename, in.pos.Line, in.pos.LineRune, e)
			}
		}
	}()

	// Invoke the parser.
	in.parseFile()
	if in.parseError != nil {
		return nil, in.parseError
	}
	in.file.Name = in.filename

	// Assign comments to nearby syntax.
	in.assignComments()

	return in.file, nil
}

// Error is called to report an error.
// The reason s is often "syntax error".
// Error does not return: it panics.
func (in *input) Error(s string) {
	if s == "syntax error" && in.lastToken != "" {
		s += " near " + in.lastToken
	}
	in.parseError = fmt.Errorf("%s:%d:%d: %v", in.filename, in.pos.Line, in.pos.LineRune, s)
	panic(in.parseError)
}

// eof reports whether the input has reached end of file.
func (in *input) eof() bool {
	return len(in.remaining) == 0
}

// peekRune returns the next rune in the input without consuming it.
func (in *input) peekRune() int {
	if len(in.remaining) == 0 {
		return 0
	}
	r,