Skip to content

Latest commit

 

History

History
429 lines (250 loc) · 10.9 KB

CHANGELOG.md

File metadata and controls

429 lines (250 loc) · 10.9 KB
Raw

0.16.0

BACKWARDS INCOMPATIBILITIES:

  • Remove Value() from the parsley.Node interface{}, differentiate between literal and non-literal nodes

0.15.0

BACKWARDS INCOMPATIBILITIES:

  • Replace Type() string on parsley.Node with Schema() interface{} to support any user-defined schema
  • Allow to set schema for nil parser

IMPROVEMENTS:

  • Use Go 1.16
  • Move tools dependency file to a subpackage

OTHER:

  • Remove CodeCov integration

0.14.0

BACKWARDS INCOMPATIBILITIES:

  • Remove NL and Whitespaces parsers, these are achievable using the trim parsers
  • Change Trim parsers to always skip all whitespaces and return specific whitespace error
  • In text.LeftTrim prefer a not found parse error over a whitespace error if the position is the same

IMPROVEMENTS:

  • Add struct wrapper for parser functions to allow recursive parser definitions
  • Simplify parsley.Error, remove unnecessary msg field (and therefore unnecessary .Error()
  • Introduce notfound errors when a parser doesn't match
  • Only collect notfound errors in any/choice if they are not in the starting position

## 0.13.4

IMPROVEMENTS:

  • Allow WsSpacesForceNl mode in RightTrim

## 0.13.3

IMPROVEMENTS:

  • Make the parsley.StaticCheck method public

## 0.13.2

BUGFIXES:

  • The Select interpreter should not run static checking on any child nodes, but simply return the index node's type

## 0.13.1

IMPROVEMENTS:

  • The Select interpreter should static check the selected node (if the node is checkable)

0.13.0

BACKWARDS INCOMPATIBILITIES:

  • The parsley.Evaluate function doesn't expect an evaluation context anymore, which was renamed to user context to make it more clear and can be set on parsley.Context
  • Move ast.WalkNode to parsley.Walk
  • The transform functions don't expect a NodeTransformer object anymore (the node transformation must be set on the interpreter)

IMPROVEMENTS:

  • The evaluation context is called user context now and can be set on parsley.Context
  • Proper recursive static checking (depth-first)
  • Proper recursive node transformation support (breadth-first)
  • The static checking and transformation are disabled by default and can be enabled on parsley.Context
  • Add NodeTransformerRegistry interface

0.12.3

IMPROVEMENTS:

  • Add NodeTransformer to context to allow AST transformations
  • The StaticChecker helper should not parse the input (minor backward incompatibility)

0.12.2

BUGFIXES:

  • Do not panic in StaticCheck if there is no interpreter

0.12.1

IMPROVEMENTS:

  • The text types should be constants (e.g. terminal.StringType)

0.12.0

IMPROVEMENTS:

  • Add parsley.NonTerminalNode interface for nonterminal nodes
  • Add parsley.StaticChecker and parsley.StaticCheckable interface for doing static analysis
  • Add type to the AST nodes

BACKWARDS INCOMPATIBILITIES:

  • The parsley.Interpreter's Eval now expects a parsley.NonTerminalNode instead of a list of nodes
  • The ast.NodeList's node methods will run always on the first node only
  • The ast.NilNode was renamed to ast.EmptyNode to avoid confusion
  • The ast.NewTerminalNode now expects a new valueType parameter
  • Most of the text/terminal parsers now will return a custom node type (string node, int node, etc.)

0.11.8

IMPROVEMENTS:

  • Remove the index from the walk function, make it properly recursive

0.11.7

IMPROVEMENTS:

  • Make the whole AST walkable, add Walk function to ast.NonTerminalNode

0.11.6

IMPROVEMENTS:

  • Add keyword handling to parse context

0.11.5

IMPROVEMENTS:

  • Add TimeDuration parser

0.11.4

IMPROVEMENTS:

  • Change Sequence.Name() to return *combinator.Sequence
  • Add Sequence.Token() method to set result token

0.11.3

IMPROVEMENTS:

  • Add new whitespace mode: text.WsSpacesForceNl to force new lines

0.11.2

IMPROVEMENTS:

  • Add text NL parser to match new lines

0.11.1

IMPROVEMENTS:

  • Allow higher quality errors to be returned by the parsers

0.11.0

BACKWARDS INCOMPATIBILITIES:

  • Rename parser's ReturnError method to Name
  • Rename Seq combinator to SeqOf
  • rename Recursive combinator to Seq (the name Recursive doesn't mean anything)

IMPROVEMENTS:

  • Add Name() function to combinator.Sequence (same helper as ReturnError)

0.10.2

IMPROVEMENTS:

  • Add SuppressError combinato

0.10.1

IMPROVEMENTS:

  • Record possible errors in Context if a combinator returns a result for better error messages

0.10.0

BACKWARDS INCOMPATIBILITIES:

  • Rename Substring parser to Op

0.9.0

BACKWARDS INCOMPATIBILITIES:

  • The parser interface was changed to return an error. The error return value was previously removed but since I realised it was a bad design decision to save ~10% in benchmarks.
  • The parser interface doesn't contain the Name() method anymore, it wasn't that useful.
  • The error field was removed from the parsing context
  • The parsing context now carries the file set
  • The top level parsley.Parse and parsley.Evaluate now will return with a simple error type which already contains the error position (these are only convenience methods).
  • I removed the *OrValue combinators as you can force a single child result with the Single combinator
  • Clean up all combinators to only have parser parameters
  • Add parser.ReturnError combinator to be able to override a parser's error (if the error's position is the same as the reader's)
  • Add ReturnError() helper method to parser.Func

BUGFIXES:

  • The string parser won't accept new-line characters (\r, \n) in double-quoted strings anymore

## 0.8.4

BUGFIXES:

  • The Single combinator will now return with the original parser's name

## 0.8.3

IMPROVEMENTS:

  • Add Go module definition
  • Update CircleCI config to use Go 1.11

0.8.2

IMPROVEMENTS:

  • Add Single combinator: Single will change the result of p if it returns with a non terminal node with only one child. In this case directly the child will be returned.

0.8.1

IMPROVEMENTS:

  • Add back the SeqTryOrValue, SepByOrValue and SepByOrValue1 combinators as these are required

BUGFIXES:

  • Many1 won't match for zero p matches

0.8.0

IMPROVEMENTS:

  • the library got a huge performance increase by refactoring the String terminal parser. The issue was caused by string concatenation and the excessive use of strconv.UnquoteChar.

BACKWARDS INCOMPATIBILITIES:

  • a new parsing context struct was introduced which carries the reader, the result cache, the call count and the best error
  • the parser interface was changed to use the parsing context
  • the parsley.History interface and the parser.History struct was replaced by parsley.ResultCache (which lives in the context now)
  • the terminal.Integer parser will return an int64

0.7.0

BACKWARDS INCOMPATIBILITIES:

  • extend parsley.Error to have a Cause() error method
  • rename parsley.NewError() to parsley.NewErrorf()
  • add parsley.NewError with error type as input
  • parsley.WrapError will store the original cause but update the error message

0.6.0

BACKWARDS INCOMPATIBILITIES:

  • major refactor of most of the API
  • most of the interfaces were moved to the parsley package
  • the position handling was rewritten similar to go's token.Pos
  • the builders were completely removed
  • the reader doesn't handle whitespaces anymore
  • the parsers' return value were simplified
  • nil node values are not allowed anymore
  • most of the combinators API's were simplified to avoid repetition (like name + token)

IMPROVEMENTS:

  • the reader became stateless
  • most of the tests were rewritten using Ginkgo/Gomega
  • common interpreters were added (array, object)
  • whitespaces can be handled precisely with new parsers (text.LeftTrim/RightTrim/Trim)
  • new error type with position (parsley.Pos)
  • new empty node type
  • new file and fileset types were introduced to support parsing multiple files better

BUGFIXES:

  • the History was using the wrong key when checking left-recursion and wasn't curtailing properly

TODO:

  • some of the old tests in the combinator package needs to be rewritten using Gingko (this means we miss a lot of test coverage)

0.5.0

BACKWARDS INCOMPATIBILITIES:

  • text.NewReader now expects a filename parameter

IMPROVEMENTS:

  • Windows-style line endings (\r\n) are automatically replaced to Unix-style line endings (\n) in the text reader.

0.4.0

BACKWARDS INCOMPATIBILITIES:

  • Typo fix in some methods: peak -> peek

IMPROVEMENTS:

  • Move precompiled whitespace regexp to a separate variable in text reader

OTHER:

  • Fix example JSON parser + add comparison benchmarks against encoding/json

0.3.3

IMPROVEMENTS:

  • allow nil pos in reader errors, replace {{err}} placeholder in WrapError
  • add filename to position, add new file reader constructor

0.3.2

IMPROVEMENTS:

  • reader.WrapError() falls back to the cause's error message if no error message was given and the cause is not a reader error

0.3.1

IMPROVEMENTS:

  • Add SepByOrValue and SepByOrValue1 combinators which will return the value node if only the value parser is matched

0.3.0

BACKWARDS INCOMPATIBILITIES:

  • Add history object to parser.Parse
  • Move Memoize back to the combinators
  • Move parsley package to a directory
  • Change top-level Parse/Evaluate methods
  • Remove parser.Stat, collect call statistics in history

IMPROVEMENTS:

  • Add Sentence root parser

0.2.3

IMPROVEMENTS:

  • reader.WrapError keeps the original error message if empty string is given as message

0.2.2

IMPROVEMENTS:

  • Add parser.FuncFactory interface

0.2.1

BUGFIXES:

  • the ast.Node mock wasn't regenerated

IMPROVEMENTS:

  • Generate mock for parser.Parser

0.2.0

BACKWARDS INCOMPATIBILITIES:

  • combinator.Memoize was removed. Use the Memoize method on the history object instead.
  • parser.Error, parser.NewError and parser.WrapError was moved to reader
  • Interpreter.Eval now returns a reader.Error instead of a general error
  • Node.Value now returns a reader.Error instead of a general error
  • Terminal.Value now returns a reader.Error instead of a general error
  • NonTerminal.Value now returns a reader.Error instead of a general error

0.1.5

IMPROVEMENTS:

  • Add cause to reader.Error, add parser.WrapError constructor

0.1.4

IMPROVEMENTS:

  • All combinators will return with a parser.Func type (not the parser.Parser interface)
  • Change internal sepBy to a factory
  • Define dependencies with Go Dep
  • Generate mocks for testing

0.1.3

DEPRECATED:

  • combinator.Memoize will be removed in version 0.2. Use the Memoize method on the history object instead.

IMPROVEMENTS:

  • You don't need to use meaningless names for memoization anymore, but you have to be careful to call Memoize for your parsers only once.

CHANGES:

  • History.GetParserIndex was removed as only the original combinator.Memoize needed it

0.1.2

IMPROVEMENTS:

  • Add copyright and license headers to all .go files
  • Improve code coverage, add notes for (hopefully) impossible panics
  • Add codecov.io integration

0.1.1

BUG FIXES:

  • IsEOF() in text.Reader was not ignoring whitespaces when ignoreWhitespaces was set to true.

IMPROVEMENTS:

  • Add Reset() method to History to allow to reuse the parsers for multiple inputs

0.1.0

First release