Parser sometimes gives a lot of strange errors, see below. When I parse file using 1 goroutine then it works just fine.

Function that does the parsing does not rely on any global state:

func parse(filename string) {
	fp, err := os.Open(filename)
	if err != nil {
		log.Fatalf("Could not open file %s: %s", filename, err.Error())
	}

	defer fp.Close()

	var b bytes.Buffer

	conv := transform.NewReader(fp, charmap.Windows1251.NewDecoder())
	parser := php7.NewParser(io.TeeReader(conv, &b), filename)
	parser.Parse()

	for _, e := range parser.GetErrors() {
		fmt.Printf("ERROR: parsing %s: %s", filename, e)
	}

	rootNode := parser.GetRootNode()

	if rootNode == nil {
		log.Printf("Could not parse %s at all due to errors", filename)
		return
	}

	rootNode.Walk(&rootWalker{
		w:         os.Stdout,
		filename:  filename,
		comments:  parser.GetComments(),
		positions: parser.GetPositions(),
		lines:     bytes.Split(b.Bytes(), []byte("\n")),
	})
}

Errors example:

syntax error: unexpected T_ENCAPSED_AND_WHITESPACE at line 409
syntax error: unexpected '}' at line 480
syntax error: unexpected T_STRING, expecting T_VARIABLE or T_ENCAPSED_AND_WHITESPACE or T_DOLLAR_
OPEN_CURLY_BRACES or T_CURLY_OPEN at line 605
...

The parser is not as performant as it could be (PHP7+ also creates AST in order to execute file):

$ php -n -r '$start = microtime(true); require("some_big_file.php"); echo "time: " . (microtime(true) - $start) . " sec\n";'
time: 0.0185 sec

$ go run test.go some_big_file.php
Errors count:  0
Parser time: 167.636796ms

test.go.zip

I made a few (very hacky) patches that significantly reduce allocations count for some critical parts and it reduced parsing time by a factor of 1.5:

# after patches
$ go run test.go some_big_file.php
Errors count:  0
Parser time: 109.522377ms

There are still plenty more allocations that profiler shows, my patches are just proof-of-concept. speedup.patch.txt

Proposal: add WithParens in analogy with func (l *Parser) WithFreeFloating(). That function asks parser to preserve info about parenthesis.

It can either introduce AST nodes that represent parenthesis or make them implicit, but positions info should respect them.

It's almost impossible to handle parenthesis in the source code right now. We can't reliably count a number of expression surrounding ( and ).

Imagine this binary expression:

(($x) + ((($y))))

If we use GetPosition and take a slice of src[pos.StartPos-1 : pos.EndPos], result would be:

$x) + ((($y

If we go to the left and right, we'll count 2 parens, but the binary expression itself has only 1 surrounding, while $y is surrounded by 3 pairs of ().

With proposed feature, I would expect binary expression position to capture the entire (($x) + ((($y)))). Sub-expression 1 should be ($x) and the 2 is ((($y))).

For clone expr like clone ($x) we get positions that encloses clone ($x.

I'm having issues with the namespace resolver. It contains unresolved names like void, true and null. Shouldn't these be removed when they are not resolved

test.php

<?php

declare(strict_types=1);

namespace App\Domain\Handler\Cart;

use SimpleBus\Message\Recorder\RecordsMessages;
use App\Domain\Command\ChangeCurrencyCommand;
use App\Domain\Repository\CartRepository;
use App\Domain\Event\CurrencyChangedEvent as CurrencyChangedEventWithAlias;

class ChangeCurrencyHandler
{
    /**
     * @var CartRepository
     */
    private $cartRepository;

    /**
     * @var RecordsMessages
     */
    private $eventRecorder;

    public function __construct(
        CartRepository $cartRepository,
        RecordsMessages $eventRecorder
    ) {
        $this->cartRepository   = $cartRepository;
        $this->eventRecorder    = $eventRecorder;
    }

    public function __invoke(ChangeCurrencyCommand $command) : void
    {
        if (true === $command->getBool()) {
            // Do something
        }
        
        if (null !== $command->getNull()) {
            // Do something
        }

        $this->eventRecorder->record(new CurrencyChangedEventWithAlias());
    }
}

main.go

package main

import (
	"fmt"
	"github.com/z7zmey/php-parser/php7"
	"github.com/z7zmey/php-parser/visitor"
	"os"
	"reflect"
)

func main() {
	for _, file := range os.Args[1:] {
		fmt.Printf("Checking %s\n", file)

		checkFile(file)
	}
}

func checkFile(file string) {
	src, err := os.Open(file)
	if err != nil {
		panic(err)
	}

	parser := php7.NewParser(src, file)
	parser.Parse()

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

	nsResolver := visitor.NewNamespaceResolver()
	parser.GetRootNode().Walk(nsResolver)

	for n, fqcn := range nsResolver.ResolvedNames {
		fmt.Printf("Found %s: %s\n", reflect.TypeOf(n), fqcn)
	}
}

output

Checking ./test.php
Found *name.Name: SimpleBus\Message\Recorder\RecordsMessages
Found *name.Name: App\Domain\Command\ChangeCurrencyCommand
Found *name.Name: void
Found *name.Name: true
Found *name.Name: null
Found *name.Name: App\Domain\Event\CurrencyChangedEvent
Found *stmt.Class: App\Domain\Handler\Cart\ChangeCurrencyHandler
Found *name.Name: App\Domain\Repository\CartRepository

I wanted to call concurrently the php7 parser but as rootnode, comments and positions are defined as php7 module variable it does not work. What I did to solve this issue, was to move those structures to the lexer struct https://github.com/z7zmey/php-parser/blob/master/scanner/lexer.go#L438 and update the yacc parser file and it has worked.

Did I miss something somewhere to have it working with goroutines ?

If not, do you have a different idea to handle such a case or would you consider a pull request ?

This is a really interesting project, thanks for working on it. I've run into what looks like an erroneous parse failure with regard to string interpolation code.

<?php
$filename = "something.txt";
@header("Content-Disposition: attachment; filename=\"$filename\"");

This fails due to parse errors of various descriptions, depending on what the surrounding code looks like. When using the php-parser binary, this dumps out

$ php-parser /tmp/brokenparse.php
==> /private/tmp/brokenparse.php
syntax error: unexpected $end, expecting ')'
  | [*stmt.StmtList]
  |   "Stmts":

PHP itself doesn't complain about this code and parses it just fine. Running php -l /tmp/brokenparse.php results in no errors. I believe this is related to a flaw in the grammar, because if $filename is followed by a space things work fine:

<?php
$filename = "something.txt";
@header("Content-Disposition: attachment; filename=\"$filename \"");

➜  analyze php-parser /tmp/brokenparse.php
==> /private/tmp/brokenparse.php
  | [*stmt.StmtList]
  |   "Position": Pos{Line: 3-4 Pos: 8-104};
  |   "Stmts":
  |     [*stmt.Expression]

... 52 lines elided ...

At some point, php parser started to parse [1, 2,] as array of 3 elements where the last array element consisting of key=nil and val=nil.

If we dig further, it also parses something like [,,,] as an array of 4 (empty) elements.

Is it intended behavior?

fixes #56

halt_compiler needs to be pulled out to the top level for this so return 0 doesn't skip unwinding the stack

__halt_compiler() may only be used in the outermost scope, so this removes any reference to in within blocks, and pulls it out of the function in the tests.

What I expected When using the package simply labelled `printer*, I assumed it would just print the file with all the formatting it had previously, however it seems this is just meant to be a pretty printer.

What's the plan for retaining formatting, if any? My use case is that I'd like to make something that'll resolve all my PHP namespaces in Sublime Text when you hit save.

For saving the data back out, I'm sure I could do a sort of hack where I only modify the lines affected, but I'm hoping retention of formatting is do-able and not too difficult so I can avoid that effort

At the very least, can the Printer struct perhaps just be renamed to PrettyPrinter?

I'm on (go version go1.14 linux/amd64)

go get github.com/z7zmey/php-parser

Shows

# github.com/z7zmey/php-parser
runtime.main_main·f: function main is undeclared in the main package

Found a problem with the following code using the dev branch:

<?php

declare(strict_types=1);

$a = "JSON_MERGE('{\"vat\"}')";

results in:

syntax error: unexpected T_STRING at line 5