Don't try to keep backwards-compatibility with the old factory
style, which doesn't map cleanly onto supported options (we only
have ONLY_PHP7/PREFER_PHP7, which should probably create a Php8
parser in terms of how they are used, but this would no longer
match their names).
Instead, I have backported the new createForNewestSupportedVersion()
and createForHostVersion() methods to PHP-Parser 4.
This doesn't make a lot of sense now that Lexer::tokenize() returns
the tokens.
The tokens for the last parse should be fetched via
Parser::getTokens() instead.
Types omitted in two places where we violate them currently:
Namespace_::$stmts can be null during parsing, and Enum_::$scalarType
can be a complex type for invalid programs.
The Lexer now only provides the tokens to the parser, while the
parser is responsible for determining which attributes are placed
on notes. This only needs to be done when the attributes are
actually needed, rather than for all tokens.
This removes the usedAttributes lexer option (and lexer options
entirely). The attributes are now enabled unconditionally. They
have less overhead now, and the need to explicitly enable them for
some use cases (e.g. formatting-preserving printing) doesn't seem
like a good tradeoff anymore.
There are some additional changes to the Lexer interface that
should be done after this, and the docs / upgrading guide haven't
been adjusted yet.
In most circumstances we are interested in the whole string, not
the parts split by namespace separator. As names are common, this
representation measurably improves memory usage and performance.
In preparation for switching this to a plain string in
PHP-Parser 5, deprecate direct access to the property and
provide an API that will work on both versions.
(cherry picked from commit c9e5a13d68486e9fd75f9be1b4639644e54e7f4f)
Normalize CRLF to LF in getReformattedText(). That was, if the
comment is pretty-printed, we will use proper LF newlines, rather
than inserting indentation between the CR and LF.
At the same time, this also makes it easier to emit actual CRLF
newlines with a custom pretty printer.
Fixes#599.
Node visitation is now properly nested. The call sequence will
now be
$visitor1->enterNode($n);
$visitor2->enterNode($n);
$visitor2->leaveNode($n);
$visitor1->leaveNode($n);
rather than
$visitor1->enterNode($n);
$visitor2->enterNode($n);
$visitor1->leaveNode($n);
$visitor2->leaveNode($n);
Fixes#899.
For types the use of a string is ambiguous -- it could be either
an Identifier or a Name. Don't guess.
Retain the implicit promotion to Identifier in places where only
Identifier is legal, e.g. various symbol names.
By making flags on the Param builder configurable by providing make(Public|Protected|Private) methods we can promote parameters to properties from the constructor
Doc strings have a trailing \n and these will get interpreted as
\r\n and removed from the string contents.
For nowdoc, fall back to single quote if there's a trailing \r.
For heredoc, escape all isolated \r -- unlike \n and \r\n this is
really a special character, because this is no longer relevant as
an actual newline character.
When detecting whether the string contains the end label, allow
leading whitespace in front of it. This is legal since the
introduction of flexible doc strings.
With the introduction of flexible doc strings, the ending label
is no longer required to be followed by a semicolon or newline.
We need to prevent doc string printing if the label is followed
by any non-label character.
This makes pretty printing round trip to another Float literal,
rather than a constant lookup. The 1e1000 form in particular is
chosen because that seems to be the typical form used in various
tests.
This makes us match the PHP 8.2 handling of readonly. Handling of
"readonly" functions is moved to the parser to allow distinguishing
them from readonly properties with DNF types. We have to uglify the
grammar to avoid some shift/reduce conflicts. Thank you WordPress.