Releases: pegjs/pegjs
0.10.0
Major Changes
-
Parsers can be generated in multiple module formats. The available
formats are: CommonJS (the default), AMD, UMD, globals, and bare (not
available from the command-line).The format can be specified using the
format
option of thepeg.generate
function or the--format
option on the command-line.It is also possible to specify parser dependencies using the
dependencies
option of thepeg.generate
function or the--dependency
/-d
option on
the command-line. This is mainly useful for the UMD format, where the
dependencies are translated into both AMD dependencies and CommonJS
require
calls. -
Browser version of PEG.js is now in the UMD format. This means it will try
to detect AMD or Node.js/CommonJS module loader and define itself as a
module. If no loader is found, it will export itself using a global
variable. -
API polishing. The
peg.buildParser
function was renamed to
peg.generate
. The global variable the browser version of PEG.js is
available in when no loader is detected was renamed fromPEG
topeg
. -
CLI improvements. There is new
--output
/-o
command-line option which
allows to specify the output file. The old way of specifying the output file
using a second argument was removed. To make room for the new-o
option
the old one (a shortcut for--optimize
) was renamed to-O
. All these
changes make PEG.js conform to traditional compiler command-line interface.It is now also possible to use
-
as a file name on the command-line (with
the usual meaning of standard input/output). -
Improved error messages. Both messages produced by PEG.js and generated
parsers were improved. -
Duplicate rule definitions are reported as errors.
-
Duplicate labels are reported as errors.
Minor Changes
- Exposed the AST node visitor builder as
peg.compiler.visitor
. This is
useful mainly for plugins which manipulate the AST. - Exposed the function which builds messages of exceptions produced by
generated parsers asSyntaxError.buildMessage
. This is useful mainly for
customizing these error messages. - The
error
andexpected
functions now accept an optionallocation
parameter. This allows to customize the location in which the resulting
syntax error is reported. - Refactored expectations reported in the
expected
property of exceptions
produced by generated parsers. They are no longer de-duplicated and sorted,
their format changed to be more machine-readable, and they no longer contain
human-readable descriptions. - The
found
property of exceptions produced by theerror
function is now
set tonull
. - Removed access to the parser object in parser code via the
parser
variable. - Made handling of optional parameters consistent with ES 2015. Specifically,
passingundefined
as a parameter value is now equivalent to not passing
the parameter at all. - Renamed several compiler passes.
- Generated parsers no longer consider
\r
,\u2028
, and\u2029
as
newlines (only\n
and\r\n
). - Simplified the arithmetics example grammar.
- Switched from
first
/rest
tohead
/tail
in PEG.js grammar and example
grammars. - Started using ESLint instead of JSHint and fixed various problems it found.
- Added contribution
guidelines. - Removed support for io.js.
Bug Fixes
- Fixed
bin/pegjs
so that invoking it with one non-option argument which is
an extension-less file doesn’t cause that file to be overwritten. - Fixed label scoping so that labels in expressions like
(a:"a")
or(a:"a" b:"b" c:"c")
aren’t visible from the outside. - Fixed escaping of generated JavaScript strings & regexps to also escape DEL
(U+007F). - Fixed the JSON example grammar to correctly handle characters with code
points above U+10FF in strings. - Fixed multiple compatibility issues of
tools/impact
on OS X. - Fixed slow deduplication of expectation descriptions.
0.9.0
Major Changes
- Tracing support. Parsers can be compiled with support for tracing their
progress, which can help debugging complex grammars. This feature is
experimental and is expected to evolve over time as experience is gained.
More details - Infinite loop detection. Grammar constructs that could cause infinite
loops in generated parsers are detected during compilation and cause errors. - Improved location information API. The
line
,column
, andoffset
functions available in parser code were replaced by a singlelocation
function which returns an object describing the current location. Similarly,
theline
,column
, andoffset
properties of exceptions were replaced by
a singlelocation
property. The location is no longer a single point but a
character range, which is meaningful mainly in actions where the range
covers action’s expression.
More details - Improved error reporting. All exceptions thrown when generating a parser
have associated location information. And all exceptions thrown by generated
parser and PEG.js itself have a stack trace (thestack
property) in
environments that supportError.captureStackTrace
. - Strict mode code. All PEG.js and generated parser code is written using
JavaScript strict mode.
Minor Changes
- Labels behave like block-scoped variables. This means parser code can see
labels defined outside expressions containing code. - Empty sequences are no longer allowed.
- Label names can’t be JavaScript reserved words.
- Rule and label names can contain Unicode characters like in JavaScript.
- Rules have to be separated either by
;
or a newline (until now, any
whitespace was enough). - The PEG.js grammar and all the example grammars were completely rewritten.
This rewrite included a number of cleanups, formatting changes, naming
changes, and bug fixes. - The parser object can now be accessed as
parser
in parser code. - Location information computation is faster.
- Added support for Node.js >= 0.10.x, io.js, and Edge. Removed support for
Node.js < 0.10.x.
Bug Fixes
- Fixed left recursion detector which missed many cases.
- Fixed escaping of U+0100—U+107F and U+1000—U+107F in generated code and
error messages. - Renamed
parse
andSyntaxError
topeg$parse
andpeg$SyntaxError
to
mark them as internal identifiers.
0.8.0
Big Changes
- Completely rewrote the code generator. Among other things, it allows optimizing generated parsers for parsing speed or code size using the
optimize
option of thePEG.buildParser
method or the--optimize
/-o
option on the command-line. All internal identifiers in generated code now also have apeg$
prefix to discourage their use and avoid conflicts. [#35, #92] - Completely redesigned error handling. Instead of returning
null
inside actions to indicate match failure, newexpected
anderror
functions can be called to trigger an error. Also, expectation inside theSyntaxError
exceptions are now structured to allow easier machine processing. [#198] - Implemented a plugin API. The list of plugins to use can be specified using the
plugins
option of thePEG.buildParser
method or the--plugin
option on the command-line. Also implemented the--extra-options
and--extra-options-file
command-line options, which are mainly useful to pass additional options to plugins. [#106] - Made
offset
,line
andcolumn
functions, not variables. They are now available in all parsers and return lazily-computed position data. Removed now uselesstrackLineAndColumn
option of thePEG.buildParser
method and the--track-line-and-column
option on the command-line. - Added a new
text
function. When called inside an action, it returns the text matched by action's expression. [#131] - Added a new
$
operator. It extracts matched strings from expressions. - The
?
operator now returnsnull
on unsuccessful match. - Predicates now always return
undefined
. - Replaced the
startRule
parameter of theparse
method in generated parsers with more genericoptions
parameter. The start rule can now be specified as thestartRule
option. Theoptions
parameter can be also used to pass custom options to the parser because it is visible as theoptions
variable inside parser code. [#37] - The list of allowed start rules of a generated parser now has to be specified explicitly using the
allowedStartRules
option of thePEG.buildParser
method or the--allowed-start-rule
option on the command-line. This will make certain optimizations like rule inlining easier in the future. - Removed the
toSource
method of generated parsers and introduced a newoutput
option of thePEG.buildParser
method. It allows callers to specify whether they want to get back the parser object or its source code. - The source code is now a valid npm package. This makes using development
versions easier. [#32] - Generated parsers are now ~25% faster and ~62%/~3% smaller (when optimized for size/speed) than those generated by 0.7.0.
- Requires Node.js 0.8.0+.
Small Changes
bin/pegjs
now outputs just the parser source if the value of the--export-var
option is empty. This makes embedding generated parsers into other files easier. [#143]- Changed the value of the
name
property ofPEG.GrammarError
instances from “PEG.GrammarError” to just “GrammarError”. This better reflects the fact that PEG.js can get required with different variable name thanPEG
. - Setup prototype chain for
PEG.GrammarError
correctly. - Setup prototype chain for
SyntaxError
in generated parsers correctly. - Fixed error messages in certain cases with trailing input [#119]
- Fixed code generated for classes starting with
\^
. [#125] - Fixed too eager proxy rules removal. [#137]
- Added a license to all vendored libraries. [#207]
- Converted the test suite from QUnit to Jasmine, cleaning it up on the way.
- Travis CI integration.
- Various internal code improvements and fixes.
- Various generated code improvements and fixes.
- Various example grammar improvements and fixes.
- Improved
README.md
. - Converted
CHANGELOG
to Markdown.