parser_builder 1.0.6

Lightweight template-based parser build system. Simple prototyping. Comfortable debugging. Effective developing.

1.0.6

• Added the ability to optimize fast parsers at the global level. This is achieved by building in two passes. During the first pass, information is collected about the use of the parsing results returned by the functions. If the return result is never used, then the parser function is built as a fast (non-resulting, void) function (on the next pass)
• Breaking change: Removed Context.contex. Instead, added the concept of the registry. Adde new metod T getRegistry<T>(owner, String name, T defaultValue) to class Context. Currently, this has an internal use when generating parsers

1.0.5

• Added and tested parser builder SeparatedListN

1.0.4

• Added and tested parser builder ManyN

1.0.3

• Added and tested parser builder BinaryExpression
• Added simple calculator parser example

1.0.2

• Fixed bug in parser builder Verify. There was an oversight in rewriting the builder implementation (from codegen based to template-based generation) for the new version. A small piece of code was omitted to generate an error with a message if verification fails with the semantic action

1.0.1

• Minor changes in README.md file
• Fixed bugs in some parser builders that generates the parsers that use (require) semantic actions even in fast mode (for verification and so on). Bugs were found with the following parser builders Satisfy, TakeWhile, TakeWhile1, Verify. This error did not result in parsing errors, but did result in the generation of uncompilable (not a valid) source code of parsers

1.0.0

• First major version

0.18.0

• After many experiments, it was decided to return everything back as it was originally. With one small exception. The build system has been greatly simplified and improved

0.17.0

• Breaking change: It was decided to simplify the implementation of parser builders (human work) and the build process (main work) by improving the generated code optimizer (optional work). These changes do not affect the definition of parsers and do not break the work. This refers to the implementation details of the parser builders. The implementation of the new code optimizer is under development as it requires the development of a simple data flow analyzer

0.16.1

• The improvement of code generation in the new version is almost completed

0.16.0

• Breaking change: Everything is implemented almost from scratch. Now code generation is done with the code generator

0.15.4

• Improved Verify parser builder. Now the failure position in ErrMessage is set not to the initial position, but to the last position. Previously, this was not possible (before the improvement of the system for working with parsing errors). This applies only to cases where verification fails. In fact, this is the purpose of the generated Verify parser
• Minor improvements to the error reporting system

0.15.3

• Another round to improve the error reporting system. Without breaking changes. This system is getting better and better. The results are already very good. All known shortcomings (and found bugs) have been fixed
• Implemented and not tested parser builder Verify. It is very, very helpful. Allows, at runtime, to influence the parsing process and to program "smart" parsing rules in the grammar. For example, define a parser that will check the number of arguments of a known function. It is programmed not flexibly, but rigidly (you can also flexibly, this is a matter of taste). The main thing is that it works. And it works on top of another parser. No need to write duplicate code. Great

0.15.2

• Fixed minor bug in parser builder SwitchTag

0.15.1

• Slightly improved the code generated by parser builder ManyTill. Generated errors are now more informative
• Another round to improve the error reporting system. Now the parsing error post processing procedures generate a more informative error report when using the builder Nested. Changes were made to improve the post-processing of ErrNested errors. Generated errors that are nested relative to an ErrNested error are now temporarily wrapped and do not affect the process of determining the furthest position of an error when filtering errors by their position. This allows more errors to be displayed and, at the same time, it allows to reduce the number of unnecessary error messages where it is required, combining them into one informative error (combination of main error and furthest errors if they are further than main error)
• Breaking change: The signature of the parser builder constructor Nested has been changed. The parameter Tag tag is non nullable now and must be specified

0.15.0

• Improved error reporting procedures. Fixed shortcomings, defects and errors in post-processing errors when creating a parser error report. Now the error messages look like they should
• Slightly changed the implementation of the experimental parser builder SwitchTag to improve error reporting
• Breaking change: The way of generating and reporting errors with nested errors has been unified. The number of error types for such cases has been reduced, instead one type has been left (ErrNested). This type has been changed and is now universal and easily customizable
• Implemented parser builder Nested (this is a reworked builder Malformed)
• Breaking change: Removed parser builder Malformed in favor of using parser builder Nested. It supports specifying an error message value and optionally specifying a tag value
• Breaking change: The Labeled parser builder has been renamed to Expected. To more accurately reflect its purpose

0.14.4

• Removed lightweight runtime. The use of this implementation did not allow a correct comparison (and filtration) of the errors in error reporting procedures. As a result, errors messages were redundant and not correct

0.14.3

• Implemented, tested and documented experimental parser builder CombinedList1
• Breaking change: Changed the signature of the method Transformation.checkArguments(). Added new parameter String information. This is done to make it easier to find the location of the exception when calling this method. The common value is the transformer template or other useful information to identify the transformer. This applies only to the implementation of transformers
• Breaking change: Changed definition and behavior of transformer VarTransformer. Now it's a combination of a variable and an expression. The definition of the variable, the key of this variable in the expression template, and the expression (using this variable) are specified. Previously, it was possible to use only a variable by name, without the ability to use it as part of an expression. This helps make it easier to the creation of more flexible, faster, configurable (including in real time) parsers
• Fixed bug in transformer NotCharClass

0.14.2

• Added many improvements to the parser builder StringValue. The overall performance of parsing data with String values can be increased by up to 10%
• Documented and tested parser builder StringValue
• Implemented, tested and documented experimental parser builder SwitchTag. This parser builder is intended mainly for the implementation of fast lexers, but can be used (with a clever approach) for other purposes. The JSON parser (example) now parses slightly faster when using this builder in JSON parser definition
• Implemented constructor ExprTransformer.value(this.expression). To simplify the definition of expression transformers without parameters
• Implemented, tested and documented parser builder Skip. This builder is intended to be used in conjunction with the builder SwitchTag to avoid reparsing already parsed simple data (such as literals or punctuation)

0.14.1

• Fixed minor bug in builder SkipWhile1
• Implemented constructor ErrUnexpected.charOrEof(this.offset, String source, [int? c]). Simplified (improved) the generated code of some parsers that parse characters and select the generated error depending on the position of parsing
• Implemented constructor ErrUnexpected.charAt(this.offset, String source

0.14.0

• Breaking change: Changes have been made to the State class. The opt field has been removed. Instead, a log field has been added to indicate the need for error logging. This is an internal change that does not affect parser definitions

0.12.2

• Implemented experimental parser builder TakeWhile1Fold
• Improved character and string parser builders. The overall performance of the parsing process can be increased up to 3%
• Improved transformers (character predicates) CharClass, NotCharClass The overall performance of the parsing process can be increased up to 2%

0.12.1

• Changed helper function signature fastBuild(). Added named parameter Map<String, Named> publish = const {}. Example of specifying arguments: publish: {'parse': _json}. This will create the function dynamic parse(String source). This makes it possible not to write public parsing functions by hand. The specified public parsing functions will be created automatically with the specified names and will be available for impossibility by users

0.12.0

• Breaking change: Transformers can now have multiple parameters. This allows them to be used as converter (mapper) functions with multiple parameters. An example of a typical use is use of them with the Map2, Map3 (etc) parsers
• Implemented and tested parser builders FoldMany0
• Implemented and tested parser builders Map2, Map3, Map4, Map5, Map6, Map7
• Breaking change: The parser builder Map$ has been renamed to Map1

0.11.0

• Breaking change: Transformers have become more powerful, much easier to use and clearer. And now they are full citizens, they have access to the building context and can allocate local variables and have access to global declarations. The number of methods has been reduced to two methods: declare (optional) and invoke. This makes it possible to implement more efficient configurable (on startup) and context-sensitive parsers
• Improved generated source code of some character and string parsers. The overall performance of the parsing process can be increased up to 3%

0.10.2

• Improved character and string parser builders. Performance can be increased up to 5-7%
• Slightly improved the code generated by transformers CharClass, NotCharClass for small character ranges. The overall performance can be increased not less than 2% due to increasing performance of the frequently called critical parser functions that parse string data (character by character) in a loop using these transformers as predicates

0.10.1

• Fixed a bug in the transformer NotCharClass that could affect the work of parsers that are sensitive to parsing 32-bit characters, to which this transformer will be passed as a predicate. A rare case, but still a bug. The transformer NotCharClass now always returns true when calling its method has32BitChars

0.10.0

• Breaking change: Another round of improvements. Now it's time to improve transformers. Transformers now allow you to solve different problems. Transformers now have three main features: declare, inline and invoke. Each transformer must support the declare and invoke capability. Support for inline is optional, but it's the support for this feature that makes them very performant (and parsers too). The most requested transformers are already available, but nothing prevents you from creating your own transformer with the necessary functionality. All parser builders have already been updated to use the new transformers, which increases performance by at least 10% and makes the source code a little smaller
• Removed all existing transformers
• Added transformers ClosureTransformer, ExprTransformer, FuncExprTransformer, FuncTransformer, VarTransformer

0.9.2

• Breaking change: The parser builder Skip has been renamed to Sequence
• Breaking change: The parser builder SkipMany0 has been removed so as not to make modifications with minor differences. Instead, it is recommended to use the following combination of parsers Many0Count(Sequence(p1, p2, etc))

0.9.1

• Minor changes in the parser builders

0.9.0

• Introduced a new feature called optional mode. Does not make breaking changes. Operates automatically. Principle of operation (briefly): tells parsers not to generate errors in optional mode because errors are not used (ignored) in this mode. Benefits: significant reduction in memory usage, 5-10% increase in performance

0.8.2

• Documented and tested parser builder EscapeSequence
• Minor improvements in the Number parser. Now the generated parser parses even a little faster. The parser is currently only used internally, when generating a JSON parser example

0.8.1

• Improved most string parser builders. The generated code is fast again
• Minor improvements in parser builders Alpha0, Alpha1, Alphanumeric0, Alphanumeric1, Digit0, Digit1, HexDigit0, HexDigit1
• Removed parser builders Crlf, LineEnding in favor of using parser builder Tags which generates code more efficiently than the original code for these removed builders

0.8.0

• Breaking change: String parser builders have been revert back. The reason is that Dart is completely unpredictable in terms of predicting performance when writing code. Performance can be reduced by 20% with just a few changes to the source code. Dart is not C/C++, what should be fast in C/C++ will not necessarily be fast in Dart

0.7.2

• Minor changes in the parser builders
• Slightly improved generated code (for short tags) by parser builders NoneOfTags, Tags. Now tags can be specified in any order, the parsing will be done correctly in every case

0.7.1

• Fixed bugs in parser builders that generate parsers use other parsers and, at the same time, restore the parsing state. Generated parsers did not restore the value of the newly introduced field of the current character state.ch. These builders include Not, Peek
• Slightly improved generated code by parser builders NoneOf, OneOf. The generated parser code has become a little faster and a little smaller in size

0.7.0

• Breaking change: String parser builders have been reimplemented to use methods state.getChar(), state.nextChar(), state.readChar() and field state.ch. The methods are implemented as static extension methods for the State<String> class. This change generates a more readable parser source code and reduces its size by 10% and may slightly reduce parser performance by 5-10%.

0.6.0

• Breaking change: Unified declaration and usage of the Transformer. The transformer code must be a function body or a callable object. If the code is the body of a function, then the function will be declared, otherwise the variable will be declared. In both cases, the generated code must be used as a function call.
• Breaking change: Removed bool search parameter from class CharClass, added RangeProcessing processing parameter

0.5.2

• Implemented and tested parser builders NoneOfEx, TagEx

0.5.1-3.dev

• Minor code changes

0.5.1-2.dev

• Implemented and tested parser builders TakeUntil, TakeUntil1
• Breaking change: Removed parser builders TakeWhileUntil, TakeWhileExcept

0.5.1-1.dev

• Lightweight versions of some classes have been added to decrease (by 2 kilobytes) the size of the runtime code
• Changed method signature ParseRuntime.getClasses(). Added named parameter bool lightweightRuntime = true
• Changed function signature fastBuild(). Added named parameter bool lightweightRuntime = true
• The parser builder Many0 has been optimized for size
• The parser builder Many1 has been optimized for size
• The parser builder SeparatedList0 has been optimized for size
• The parser builder SeparatedList1 has been optimized for size

0.5.0

• Breaking change: The Err type has been redesigned to reduce the size of the memory consumption. Now error generation consumes 20-25% less memory

0.4.5-2.dev

• Added documentation for parser builders from the library character
• Implemented and tested parser builders ManyTill
• Added some documentation for runtime classes

0.4.5-1.dev

• Optimized an operation algorithm of the transformer CharClass
• Fixed bug in transformer CharClass
• Implemented and tested parser builders SkipMany0, TakeWhileExcept

0.4.4

• The syntax used in CharClass has been extended. It is now possible to combine ranges [A-Za-z0-9_] | "$".

0.4.3-2-dev

• Fixed bug in method tagNoCase in extension on String

0.4.3-1-dev

• Implemented and tested the optional feature to use binary search for the transformer CharClass. It can be turned on or off when measuring the effectiveness of using a particular search method
• The Allocator now supports variable names
• Implemented and tested parser builders Crlf, LineEnding, TakeWhileUntil
• Added parameter bool format = true to the function fastBuild()
• The numbering algorithm for local variables has been slightly improved

0.4.2

• Implemented and tested transformer CharClass

0.4.1-1.dev

• Fixed bug in ParserBuilder.buildBodyEx()

0.4.0

• Breaking change: The generated parser body is no longer wrapped in a for statement. This makes the code smaller and more readable

0.3.1-1.dev

• Removed class Result from runtime library
• Implemented and tested parser builders Skip
• Added documentation for parser builders from the library bytes
• Breaking change: Renamed parser builder MapRes to Map$. The MapRes will be implemented later

0.3.0

• Breaking change: The generated parser no longer requires result value wrapping. This increases parsing performance by 10-20%

0.2.1-4.dev

• Added a script fast_build.dart with a function fastBuild to simplify build process
• Implemented and tested parser builders Alphanumeric0, Alphanumeric1, AnyChar, HexDigit0, HexDigit1

0.2.1-3.dev

• Implemented and tested parser builders SkipWhile1
• Added support for an unofficial way to display error messages (for development purposes) via the getErrorMessageProcessor function
• Breaking change: Removed all building methods from context class Context
• • Added class ParseRuntime with source codes of static declarations necessary for the runtime of the generated parsers

0.2.1-2.dev

• Implemented and tested parser builders Alpha0, Alpha1, Digit0, Digit1, NoneOfTags, TakeWhile1
• A new method BuildBodyEx has been introduced, which is currently the basis for building the parser body from a template
• Renamed parser builder SkipWhile0 to SkipWhile
• Breaking change: The parser builders TakeWhile, TakeWhileMN now return a parsed string instead of a character list. It was a mistake to implement them to return a list

0.2.1-1.dev

• Implemented and tested parser builders Consumed, Not, Peek, Recognize, Tags, Tuple2, Tuple3, Tuple4, Tuple5, Tuple6, Tuple7
• Changed parser builder Eof from Eof extends ParserBuilder<String, dynamic> to Eof<I> extends ParserBuilder<I, dynamic>. Now it can be used not only with String data

0.2.0

• Removed support of lookahead as it didn't provide a significant performance boost
• Breaking change: significantly redesigned (unified) and improved the way to define the parser builder. Now it is not required to call any processing methods

0.1.6

• Added support for lookahead
• Implemented and tested parser builders Many0, Many0Count, Many1, Many1Count, ManyMN, SeparatedList1, TagNoCase
• Added tests for parser builder SeparatedList0

0.1.5

• Added field Object? context to the State class (the state of the parsing) to possibility the implementation of context-sensitive parsers (by eg. YAML parsers)

0.1.4

• Parser builders have been moved to separate classes

0.1.3

• Added tests for parser builder SeparatedPair

0.1.2

• Added tests for parser builders Eof, MapRes, Opt, Value

0.1.1

• Added tests for parser builders Satisfy, SkipWhile0

0.1.0

• Initial release