Wrong line number when parsing string.

[ganlvtech]

Dev branch.

Test Script

<?php
echo "$
$
$
$
$
";
echo __LINE__;

package main

import (
	"fmt"
	"os"

	"github.com/z7zmey/php-parser/php5"
	"github.com/z7zmey/php-parser/visitor"
)

func main() {
	bytes := []byte(`<?php
echo "$
$
$
$
$
";
echo __LINE__;
`)

	parser := php5.NewParser(bytes)
	parser.Parse()

	for _, e := range parser.GetErrors() {
		fmt.Println(e, e.Pos)
	}

	visitor := visitor.Dumper{
		Writer: os.Stdout,
		Indent: "",
	}

	rootNode := parser.GetRootNode()
	rootNode.Walk(&visitor)
}

Expected Result

[*node.Root]
  "Position": Pos{Line: 2-8 Pos: 6-39}
  "Stmts":
    [*stmt.Echo]
      "Position": Pos{Line: 2-7 Pos: 6-24}
      "Exprs":
        [*scalar.String]
          "Position": Pos{Line: 2-7 Pos: 11-23}
          "Value": "\"$\n$\n$\n$\n$\n\""
    [*stmt.Echo]
      "Position": Pos{Line: 8-8 Pos: 25-39}
      "Exprs":
        [*scalar.MagicConstant]
          "Position": Pos{Line: 8-8 Pos: 30-38}
          "Value": "__LINE__"

Actual Result

[*node.Root]
  "Position": Pos{Line: 2-3 Pos: 6-39}
  "Stmts":
    [*stmt.Echo]
      "Position": Pos{Line: 2-2 Pos: 6-24}
      "Exprs":
        [*scalar.String]
          "Position": Pos{Line: 2-2 Pos: 11-23}
          "Value": "\"$\n$\n$\n$\n$\n\""
    [*stmt.Echo]
      "Position": Pos{Line: 3-3 Pos: 25-39}
      "Exprs":
        [*scalar.MagicConstant]
          "Position": Pos{Line: 3-3 Pos: 30-38}
          "Value": "__LINE__"

My Suggestion

php-parser/scanner/scanner.rl

Lines 68 to 104 in a7e37ad

	constant_string =
	start: (
	"'" -> qoute
	| "b"i? '"' -> double_qoute
	),
	qoute: (
	(any - [\\'\r\n]) -> qoute
	| "\r" @{if lex.p+1 != eof && lex.data[lex.p+1] != '\n' {lex.NewLines.Append(lex.p)}} -> qoute
	| "\n" @{lex.NewLines.Append(lex.p)} -> qoute
	| "\\" -> qoute_any
	| "'" -> final
	),
	qoute_any: (
	any_line -> qoute
	),
	double_qoute: (
	(any - [\\"${\r\n]) -> double_qoute
	| "\r" @{if lex.p+1 != eof && lex.data[lex.p+1] != '\n' {lex.NewLines.Append(lex.p)}} -> double_qoute
	| "\n" @{lex.NewLines.Append(lex.p)} -> double_qoute
	| "\\" -> double_qoute_any
	| '"' -> final
	| '$' -> double_qoute_nonvarname
	| '{' -> double_qoute_nondollar
	),
	double_qoute_any: (
	any_line -> double_qoute
	),
	double_qoute_nondollar: (
	'"' -> final
	| "\\" -> double_qoute_any
	| [^$\\"] -> double_qoute
	),
	double_qoute_nonvarname: (
	'"' -> final
	| "\\" -> double_qoute_any
	| /[^"\\{a-zA-Z_\x7f-\xff]/ -> double_qoute
	);

constant_string won't parse double quoted string. So maybe we don't need double_qoute_nonvarname and double_qoute_nondollar state.

[z7zmey]

The scanner must interpret double-quoted strings that don't contain variables as T_CONSTANT_ENCAPSED_STRING.
double_qoute_nonvarname and double_qoute_nondollar prevents transition to final state if variable found.