v0.23.0

Added

• More support for Prism::RipperCompat is added.
• A significantly faster offset cache for Prism::Translation::Parser is added for files with multibyte characters.
• Prism::Translation::RubyParser is added.
• Prism::ConstantPathTarget#full_name is added.
• version: "3.4.0" is added as an option that is an alias for version: "latest".
• Four new APIs are added to Prism::Location:
  • Prism::Location#start_code_units_offset
  • Prism::Location#end_code_units_offset
  • Prism::Location#start_code_units_column
  • Prism::Location#end_code_units_column
• Invalid multibyte characters are now validated within strings, lists, and heredocs.

Changed

• When defining def !@, the name_loc was previously only pointing to !, but now includes the @. The name is the same.
• Prism::RipperCompat has been moved to Prism::Translation::Ripper.
• Many of the error messages that prism produces have been changed to match the error messages that CRuby produces.

v0.22.0

Added

• More support for Prism::RipperCompat is added.
• Support for Ruby 2.7 has been added, and the minimum Ruby requirement has been lowered to 2.7.

Changed

• The error for an invalid source encoding has a new :argument level to indicate it raises an argument error.
• BeginNode nodes that are used when a class, singleton class, module, method definition, or block have an inline rescue/ensure/else now have their opening locations set to the beginning of the respective keyword.
• Improved error messages for invalid characters.
• Prism.parse_file and similar APIs will raise more appropriate errors when the file does not exist or cannot be mapped.
• Correctly handle the recover parameter for Prism::Translation::Parser.

v0.21.0

Added

• Add the pm_constant_pool_find API for finding a constant.

Changed

• Fixes for Prism::Translation::Parser.
  • Ensure all errors flow through parser.diagnostics.process.
  • Fix the find pattern node.
  • Fix block forwarding with NumberedParametersNode.
  • Ensure we can parse strings with invalid bytes for the encoding.
  • Fix hash pairs in pattern matching.
• Properly reject operator writes on operator calls, e.g., a.+ -= b.
• Fix multi-byte escapes.
• Handle missing body in begin within the receiver of a method call.

v0.20.0

Added

• String literal hash keys are now marked as frozen as static literal.
• IndexTargetNode now always has the ATTRIBUTE_WRITE flag.
• Call*Node nodes now have an IGNORE_VISIBILITY flag.
• We now support the it default parameter.
• Errors and warnings now have levels associated with them.
• Symbols now have correct encoding flags.
• We have now merged parser-prism in, which provides translation to the whitequark/parser AST.
• We now emit errors for invalid method definition receivers.

Changed

• We now emit errors on invalid pinned local variables.
• When passed scopes, it is now assumed that the innermost scope is the current binding.
• We now provide better error recovery for non terminated heredocs.
• Fix for RationalNode#value for non-decimal integers.
• Unary symbols !@ and ~@ now unescape to ! and ~, respectively.
• frozen_string_literal: false now works properly.

Removed

• We've removed the locals_body_index field.
• We've removed the verbose option on the various parse APIs. Warnings are now always emitted with their associated level so that consumers can decide how to handle them.

v0.19.0

Added

• ArrayNode now has a contains_splat? flag if it has a splatted element in it.
• All of the remaining encodings have been implemented.
• Allow forwarding & in a method that has a ... parameter.
• Many statements that are found in non-statement positions are being properly rejected now.
• Void values are now properly checked.
• Referencing a parameter in its own default value is now properly rejected.
• DATA/__END__ is now parsed as its own field on parse result (data_loc) as opposed to as a comment.
• Blank * now properly forwards into arrays.
• ImplicitRestNode is introduced to represent the implicit rest of a destructure.
• We now support negative start lines.
• StringNode#heredoc?, InterpolatedStringNode#heredoc?, XStringNode#heredoc?, and InterpolatedXStringNode#heredoc? are introduced.
• NumberedParametersNode is introduced to represent the implicit set of parameters when numbered parameters are used.
• Prism::parse_success? and Prism::parse_failure? are introduced to bypass reifying the AST.
• We now emit a warning for constant assignments in method definitions.
• We now provide flags on strings and xstrings to indicate the correct encoding.
• The hash pattern rest field now more accurately parses ** and **nil.
• The equality operators are now properly parsed as non-associative.

Changed

• BREAKING: Many fields have changed positions within their nodes. This impacts the C API and the Ruby API if you are manually creating nodes through the initializer.
• BREAKING: Almost all of the error messages have been updated to begin with lowercase characters to match ruby/spec.
• Unterminated strings with only plain content are now always StringNode as opposed to InterpolatedStringNode
• BREAKING: Call node has been split up when it is in the target position into CallTargetNode and IndexTargetNode.

v0.18.0

Added

• The ParametersNode#signature method is added, which returns the same thing as Method#parameters.
• Visitor functionality has been added to the JavaScript API.
• The Node#to_dot API has been added to convert syntax trees to Graphviz digraphs.
• IfNode and UnlessNode now have a then_keyword_loc field.
• Many more encodings are now supported.
• Some new Location APIs have been added for dealing with characters instead of bytes, which are: start_character_offset, end_character_offset, start_character_column, and end_character_column.
• A warning has been added for when END {} is used within a method.
• ConstantPathNode#full_name{,_parts} will now raise an error if the receiver of the constant path is not itself a constant.
• The in keyword and the => operator now respect non-associativity.
• The .. and ... operators now properly respect non-associativity.

Changed

• Previously ... in blocks was accepted, but it is now properly rejected.
• BREAKING: librubyparser.* has been renamed to libprism.* in the C API.
• We now properly reject floats with exponent and rational suffixes.
• We now properly reject void value expressions.
• BREAKING: The --disable-static option has been removed from the C extension.
• The rescue modifier keyword is now properly parsed in terms of precedence.
• We now properly reject defining a numbered parameter method.
• BREAKING: MatchWriteNode now has a list of targets, which are local variable target nodes. This is instead of locals which was a constant list. This is to support writing to local variables outside the current scope. It has the added benefit of providing location information for the local variable targets.
• BREAKING: CaseNode has been split into CaseNode and CaseMatchNode, the latter is used for case ... in expressions.
• BREAKING: StringConcatNode has been removed in favor of using InterpolatedStringNode as a list.

v0.17.1

Changed

• Do not use constant nesting in RBI files.

v0.17.0

Added

• We now properly support forwarding arguments into arrays, like def foo(*) = [*].
• We now have much better documentation for the C and Ruby APIs.
• We now properly provide an error message when attempting to assign to numbered parameters from within regular expression named capture groups, as in /(?<_1>)/ =~ "".

Changed

• BREAKING: KeywordParameterNode is split into OptionalKeywordParameterNode and RequiredKeywordParameterNode. RequiredKeywordParameterNode has no value field.
• BREAKING: Most of the Prism:: APIs now accept a bunch of keyword options. The options we now support are: filepath, encoding, line, frozen_string_literal, verbose, and scopes. See the pull request for more details.
• BREAKING: Comments are now split into three different classes instead of a single class, and the type field has been removed. They are: InlineComment, EmbDocComment, and DATAComment.

v0.16.0

Added

• InterpolatedMatchLastLineNode#options and MatchLastLineNode#options are added, which are the same methods as are exposed on InterpolatedRegularExpressionNode and RegularExpressionNode.
• The project can now be compiled with wasi-sdk to expose a WebAssembly interface.
• ArgumentsNode#keyword_splat? is added to indicate if the arguments node has a keyword splat.
• The C API pm_prettyprint has a much improved output which lines up closely with Node#inspect.
• Prism now ships with RBS and RBI type signatures (in the /sig and /rbi directories, respectively).
• Prism::parse_comments and Prism::parse_file_comments APIs are added to extract only the comments from the source code.

Changed

• BREAKING: Multi{Target,Write}Node#targets is split up now into lefts, rest, and rights. This is to avoid having to scan the list in the case that there are splat nodes.
• Some bugs are fixed on Multi{Target,Write}Node accidentally creating additional nesting when not necessary.
• BREAKING: RequiredDestructuredParameterNode has been removed in favor of using MultiTargetNode in those places.
• BREAKING: HashPatternNode#assocs has been renamed to HashPatternNode#elements. HashPatternNode#kwrest has been renamed to HashPatternNode#rest.

v0.15.1

Changed

• Fix compilation warning on assigning to bitfield.