CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

Unreleased

[6.1.1] - 2023-03-21

Changed

• Fixed a bug where the call chain formatter was incorrectly looking at call messages.

[6.1.0] - 2023-03-20

Added

• The stree ctags command for generating ctags like universal-ctags or ripper-tags would.
• The definedivar YARV instruction has been added to match CRuby's implementation.
• We now generate better Sorbet RBI files for the nodes in the tree and the visitors.
• SyntaxTree::Reflection.nodes now includes the visitor method.

Changed

• We now explicitly require pp in environments that need it.

6.0.2 - 2023-03-03

Added

• The WithScope visitor mixin will now additionally report local variables defined through regular expression named captures.
• The WithScope visitor mixin now properly handles destructured splat arguments in required positions.

Changed

• Fixed the AST output by adding blocks to Command and CommandCall nodes in the FieldVisitor.
• Fixed the location of lambda local variables (e.g., ->(; a) {}).

6.0.1 - 2023-02-26

Added

• The class declarations returned as the result of the indexing operation now have their superclass as a field. It is returned as an array of constants. If the superclass is anything other than a constant lookup, then it raises an error.

Changed

• The nesting field on the results of the indexing operation is no longer a single flat array. Instead it is an array of arrays, where each array is a single nesting level. This more accurately reflects the nesting of the nodes in the tree. For example, class Foo::Bar::Baz; end would result in [Foo, Bar, Baz], but that incorrectly implies that you can see constants at each of those levels. Now this would result in [[Foo, Bar, Baz]] to indicate that it can see either the top level or constants within the scope of Foo::Bar::Baz only.
• When formatting hashes that have omitted values and mixed hash rockets with labels, the formatting now maintains whichever delimiter was used in the source. This is because forcing the use of hash rockets with omitted values results in a syntax error.
• Handle the case where a bare hash is used after the break, next, or return keywords. Previously this would result in hash labels which is not valid syntax. Now it maintains the delimiters used in the source.
• The << operator will now break on chained << expressions. Previously it would always stay flat.

6.0.0 - 2023-02-10

Added

• SyntaxTree::BasicVisitor::visit_methods has been added to allow you to check multiple visit methods inside of a block. There was a method called visit_methods previously, but it was undocumented because it was meant as a private API. That method has been renamed to valid_visit_methods.
• rake sorbet:rbi has been added as a task within the repository to generate an RBI file corresponding to the nodes in the tree. This can be used to help aid consumers of Syntax Tree that are using Sorbet.
• SyntaxTree::Reflection has been added to allow you to get information about the nodes in the tree. It is not required by default, since it takes a small amount of time to parse node.rb and get all of the information.
• SyntaxTree::Node#to_mermaid has been added to allow you to generate a Mermaid diagram of the node and its children. This is useful for debugging and understanding the structure of the tree.
• SyntaxTree::Translation has been added as an experimental API to transform the Syntax Tree syntax tree into the syntax trees represented by the whitequark/parser and rubocop/rubocop-ast gems.
  • SyntaxTree::Translation.to_parser(node, buffer) will return a Parser::AST::Node object.
  • SyntaxTree::Translation.to_rubocop_ast(node, buffer) will return a RuboCop::AST::Node object.
• SyntaxTree::index and SyntaxTree::index_file have been added to allow you to get a list of all of the classes, modules, and methods defined in a given source string or file.
• Various convenience methods have been added:
  • SyntaxTree::format_file - which calls format with the result of reading the file
  • SyntaxTree::format_node - which formats the node directly
  • SyntaxTree::parse_file - which calls parse with the result of reading the file
  • SyntaxTree::search_file - which calls search with the result of reading the file
  • SyntaxTree::Node#start_char - which is the same as calling node.location.start_char
  • SyntaxTree::Node#end_char - which is the same as calling node.location.end_char
• SyntaxTree::Assoc nodes can now be formatted on their own without a parent hash node.
• SyntaxTree::BlockVar#arg0? has been added to check if a single required block parameter is present and would potentially be expanded.
• More experimental APIs have been added to the SyntaxTree::YARV module, including:
  • SyntaxTree::YARV::ControlFlowGraph
  • SyntaxTree::YARV::DataFlowGraph
  • SyntaxTree::YARV::SeaOfNodes

Changed

Major changes

• BREAKING Updates to WithEnvironment:
  • The WithEnvironment module has been renamed to WithScope.
  • The current_environment method has been renamed to current_scope.
  • The with_current_environment method has been removed.
  • Previously scopes were always able to look up the tree, as in: a = 1; def foo; a = 2; end would see only a single a variable. That has been corrected.
  • Previously accessing variables from inside of blocks that were not shadowed would mark them as being local to the block only. This has been correct.
• BREAKING Lots of constants moved out of SyntaxTree::Visitor to just SyntaxTree:
  • SyntaxTree::Visitor::FieldVisitor is now SyntaxTree::FieldVisitor
  • SyntaxTree::Visitor::JSONVisitor is now SyntaxTree::JSONVisitor
  • SyntaxTree::Visitor::MatchVisitor is now SyntaxTree::MatchVisitor
  • SyntaxTree::Visitor::MutationVisitor is now SyntaxTree::MutationVisitor
  • SyntaxTree::Visitor::PrettyPrintVisitor is now SyntaxTree::PrettyPrintVisitor
• BREAKING Lots of constants are now autoloaded instead of required by default. This is only particularly relevant if you are in a forking environment and want to preload constants before forking for better memory usage with copy-on-write.
• BREAKING The SyntaxTree::Statements#initialize method no longer accepts a parser as the first argument. It now mirrors the other nodes in that it accepts its children and location. As a result, Syntax Tree nodes are now marshalable (and therefore can be sent over DRb). Previously the Statements node was not able to be marshaled because it held a reference to the parser.

Minor changes

• Many places where embedded documents (=begin to =end) were being treated as real comments have been fixed for formatting.
• Dynamic symbols in keyword pattern matching now have better formatting.
• Endless method definitions used to have a SyntaxTree::BodyStmt node that had any kind of node as its statements field. That has been corrected to be more consistent such that now going from def_node.bodystmt.statements always returns a SyntaxTree::Statements node, which is more consistent.
• We no longer assume that fiddle is able to be required, and only require it when it is actually needed.

Tiny changes

• Empty parameter nodes within blocks now have more accurate location information.
• Pinned variables have more correct location information now. (Previously the location was just around the variable itself, but it now includes the pin.)
• Array patterns in pattern matching now have more accurate location information when they are using parentheses with a constant present.
• Find patterns in pattern matching now have more correct location information for their left and right fields.
• Lots of nodes have more correct types in the comments on their attributes.
• The expressions break foo.bar :baz do |qux| qux end and next fun foo do end now correctly parses as a control-flow statement with a method call that has a block attached, as opposed to a control-flow statement with a block attached.
• The expression self::a, b = 1, 2 would previously yield a SyntaxTree::ConstPathField node for the first element of the left-hand-side of the multiple assignment. Semantically this is incorrect, and we have fixed this to now be a SyntaxTree::Field node instead.

5.3.0 - 2023-01-26

Added

• #arity has been added to DefNode, BlockNode, and Params. The method returns a range where the lower bound is the minimum and the upper bound is the maximum number of arguments that can be used to invoke that block/method definition.
• #arity has been added to CallNode, Command, CommandCall, and VCall nodes. The method returns the number of arguments included in the invocation. For splats, double splats, or argument forwards, this method returns Float::INFINITY.
• SyntaxTree::index and SyntaxTree::index_file APIs have been added to collect a list of classes, modules, and methods defined in a given source string or file, respectively. These APIs are experimental and subject to change.
• A plugin/disable_auto_ternary plugin has been added the disables the formatted that automatically changes permissable if/else clauses into ternaries.

Changed

• Files are now only written from the CLI if the content of them changes, which should match watching files less chaotic.
• In the case that rb_iseq_load cannot be found, Fiddle::DLError is now rescued.
• Previously if there were invalid UTF-8 byte sequences after the __END__ keyword the parser could potentially have crashed when parsing comments. This has been fixed.
• Previously there was special formatting for array literals that contained only variable references (either locals, method calls, or constants). For consistency, this has been removed and all array literals are now formatted the same way.

5.2.0 - 2023-01-04

Added

• An experiment in evaluating compiled instruction sequences has been added to Syntax Tree. This is subject to change, so it will not be well documented or testing at the moment. It does not impact other functionality.

Changed

• Empty parentheses on method calls will now be left in place. Previously they were left in place if the method being called looked like a constant. Now they are left in place for all method calls since the method name can mirror the name of a local variable, in which case the parentheses are required.

5.1.0 - 2022-12-28

Added

• An experiment in working with instruction sequences has been added to Syntax Tree. This is subject to change, so it is not well documented or tested at the moment. It does not impact other functionality.
• You can now format at a different base layer of indentation. This is an optional third argument to SyntaxTree::format.

Changed

• Support forwarding anonymous keyword arguments with **.
• The BodyStmt node now has a more correct location information.
• Ignore the textDocument/documentColor request coming into the language server to support clients that require that request be received.
• Do not attempt to convert if..else into ternaries if the predicate has a Binary node.
• Properly handle nested pattern matching when a rightward assignment is inside a when clause.

5.0.1 - 2022-11-10

Changed

• Fix the plugin parsing on the CLI so that they are respected.

5.0.0 - 2022-11-09

Added

• Every node now implements the #copy(**) method, which provides a copy of the node with the given attributes replaced.
• Every node now implements the #===(other) method, which checks if the given node matches the current node for all attributes except for comments and location.
• There is a new SyntaxTree::Visitor::MutationVisitor and its convenience method SyntaxTree.mutation which can be used to mutate a syntax tree. For details on how to use this visitor, check the README.

Changed

• Nodes no longer have a comments: keyword on their initializers. By default, they initialize to an empty array. If you were previously passing comments into the initializer, you should now create the node first, then call node.comments.concat to add your comments.
• A lot of nodes have been folded into other nodes to make it easier to interact with the AST. This means that a lot of visit methods have been removed from the visitor and a lot of class definitions are no longer present. This also means that the nodes that received more function now have additional methods or fields to be able to differentiate them. Note that none of these changes have resulted in different formatting. The changes are listed below:
  • IfMod, UnlessMod, WhileMod, UntilMod have been folded into IfNode, UnlessNode, WhileNode, and UntilNode. Each of the nodes now have a modifier? method to tell if it was originally in the modifier form. Consequently, the visit_if_mod, visit_unless_mod, visit_while_mod, and visit_until_mod methods have been removed from the visitor.
  • VarAlias is no longer a node, and the Alias node has been renamed. They have been folded into the AliasNode node. The AliasNode node now has a var_alias? method to tell you if it is aliasing a global variable. Consequently, the visit_var_alias method has been removed from the visitor interface. If you were previously using this method, you should now use visit_alias instead.
  • Yield0 is no longer a node, and the Yield node has been renamed. They has been folded into the YieldNode node. The YieldNode node can now have its arguments field be nil. Consequently, the visit_yield0 method has been removed from the visitor interface. If you were previously using this method, you should now use visit_yield instead.
  • FCall is no longer a node, and the Call node has been renamed. They have been folded into the CallNode node. The CallNode node can now have its receiver and operator fields be nil. Consequently, the visit_fcall method has been removed from the visitor interface. If you were previously using this method, you should now use visit_call instead.
  • Dot2 and Dot3 are no longer nodes. Instead they have become a single new RangeNode node. This node looks the same as Dot2 and Dot3, except that it additionally has an operator field that contains the operator that created the node. Consequently, the visit_dot2 and visit_dot3 methods have been removed from the visitor interface. If you were previously using these methods, you should now use visit_range instead.
  • Def, DefEndless, and Defs have been folded into the DefNode node. The DefNode node now has the target and operator fields which originally came from Defs which can both be nil. It also now has an endless? method on it to tell if the original node was found in the endless form. Finally the bodystmt field can now either be a BodyStmt as it was or any other kind of node since that was the body of the DefEndless node. The visit_defs and visit_def_endless methods on the visitor have therefore been removed.
  • DoBlock and BraceBlock have now been folded into a BlockNode node. The BlockNode node now has a keywords? method on it that returns true if the block was constructed with the do..end keywords. The visit_do_block and visit_brace_block methods on the visitor have therefore been removed and replaced with the visit_block method.
  • Return0 is no longer a node, and the Return node has been renamed. They have been folded into the ReturnNode node. The ReturnNode node can now have its arguments field be nil. Consequently, the visit_return0 method has been removed from the visitor interface. If you were previously using this method, you should now use visit_return instead.
• The ArgsForward, Redo, Retry, and ZSuper nodes no longer have value fields associated with them (which were always string literals corresponding to the keyword being used).
• The Command and CommandCall nodes now has block attributes on them. These attributes are used in the place where you would previously have had a MethodAddBlock structure. Where before the MethodAddBlock would have the command and block as its two children, you now just have one command node with the block attribute set to the Block node.
• Previously the formatting options were defined on an unfrozen hash called SyntaxTree::Formatter::OPTIONS. It was globally mutable, which made it impossible to reference from within a Ractor. As such, it has now been replaced with SyntaxTree::Formatter::Options.new which creates a new options object instance that can be modified without impacting global state. As a part of this change, formatting can now be performed from within a non-main Ractor. In order to check if the plugin/single_quotes plugin has been loaded, check if SyntaxTree::Formatter::SINGLE_QUOTES is defined. In order to check if the plugin/trailing_comma plugin has been loaded, check if SyntaxTree::Formatter::TRAILING_COMMA is defined.

4.3.0 - 2022-10-28

Added

• #183 - Support TruffleRuby by eliminating internal pattern matching in some places and stopping some tests from running in other places.
• #184 - Remove internal pattern matching entirely.

Changed

• #183 - Pattern matching works against dynamic symbols now.
• #184 - Exit with the correct exit status within the rake tasks.

4.2.0 - 2022-10-25

Added

• #182 - The new stree expr CLI command will function similarly to the stree match CLI command except that it only outputs the first expression of the program.
• #182 - Added the SyntaxTree::Pattern class for compiling in expressions into procs.

Changed

• #182 - Much more syntax is now supported by the search command.

4.1.0 - 2022-10-24

Added

• #180 - The new stree search CLI command and the corresponding SyntaxTree::Search class for searching for a pattern against a given syntax tree.

4.0.2 - 2022-10-19

Changed

• #177 - Fix up various other issues with the environment visitor addition.

4.0.1 - 2022-10-18

Changed

• #172 - Use a refinement for Symbol#name addition so that other runtimes or tools don't get confused by its availability.
• #173 - Fix the current_environment usage to use the method instead of the instance variable.
• #175 - Update prettier_print requirement since v1.0.0 had a bug with #breakable_return.

4.0.0 - 2022-10-17

Added

• #169 - You can now pass --ignore-files multiple times.
• #157 - We now support tracking local variable definitions throughout the visitor. This allows you to access scope information while visiting the tree.
• #170 - There is now an undocumented STREE_FAST_FORMAT environment variable checked when formatting. It has the effect of turning off formatting call chains and ternaries in special ways. This improves performance quite a bit. I'm leaving it undocumented because ideally we just improve the performance as a whole. This is meant as a stopgap until we get there.

Changed

• #170 - We now require at least version 1.0.0 of prettier_print. This is to take advantage of the first-class string support in the doc tree.
• #170 - Pattern matching has been removed from usage internal to this library (excluding the language server). This should hopefully enable runtimes that don't have pattern matching fully implemented yet (e.g., TruffleRuby) to run this gem.

3.6.3 - 2022-10-11

Changed

• #167 - Change the error encountered when an else node does not have an associated end token to be a parse error.

3.6.2 - 2022-10-04

Changed

• #165 - Conditionals (if/unless), loops (for/while/until) and lambdas all had issues when comments immediately succeeded the declaration of the node where the comment could potentially be dropped. That has now been fixed.
• #166 - Blocks can now be formatted even if they are the top node of the tree. Previously they were looking to their parent for some additional metadata, so we now handle the case where the parent is nil.

3.6.1 - 2022-09-28

Changed

• #161 - Previously, we were checking if STDIN was a TTY to determine if there was content to be read. Instead, we now check if no filenames were passed, and in that case we attempt to read from STDIN. This should fix errors users were experiencing in non-TTY environments like CI.
• #162 - Parse errors shouldn't crash the language server anymore.

3.6.0 - 2022-09-19

Added

• #158 - Support the ability to pass --ignore-files to the CLI and the Rake tasks to ignore a certain pattern of files.

3.5.0 - 2022-08-26

Added

• #148 - Support Ruby 2.7.0 (previously we only supported back to 2.7.3).
• #152 - Support the -e inline script option for the stree CLI.

Changed

• #141 - Use q.format for SyntaxTree.format so that the main node gets pushed onto the stack for checking parent nodes.
• #147 - Fix rightward assignment token management such that in and => stay the same regardless of their context.

3.4.0 - 2022-08-19

Added

• #127 - Allow the language server to handle other file extensions if it is activated for those extensions.
• #133 - Add documentation on supporting vim and neovim.

Changed

• #132 - Provide better error messages when end quotes and end keywords are missing from tokens.
• #134 - Ensure the correct end keyword is getting removed by begin..rescue clauses.
• #137 - Better support regular expressions with no ending token.

3.3.0 - 2022-08-02

Added

• #123 - Allow the rake tasks to configure print width.
• #125 - Add support for an .streerc file in the current working directory to configure the CLI.

3.2.1 - 2022-07-22

Changed

• #119 - If there are conditionals in the assignment we cannot convert it to the modifier form. There was a bug where it would stop checking for assignment nodes if there were any optional child nodes.

3.2.0 - 2022-07-19

Added

• #116 - Pass the --print-width option in the CLI to the language server.

3.1.0 - 2022-07-19

Added

• #115 - Support the --print-width option in the CLI for the actions that support it.

3.0.1 - 2022-07-15

Changed

• #112 - Fix parallel CLI execution by not short-circuiting with the || operator.

3.0.0 - 2022-07-04

Changed

• #102 - Handle requests to the language server for files that do not yet exist on disk.

Removed

• #108 - Remove old inlay hints code.

2.9.0 - 2022-07-04

Added

• #106 - Add inlay hint support to match the LSP specification.

2.8.0 - 2022-06-21

Added

• #95 - The HeredocEnd node has been added which effectively results in the ability to determine the location of the ending of a heredoc from source.
• #99 - The LSP now allows you to pass the same configuration options as the other CLI commands which allows formatting to be modified in the VSCode extension.
• #100 - The LSP now explicitly responds to the shutdown request so that VSCode never deadlocks.

Changed

• #96 - The CLI now runs in parallel by default. There is a worker created for each processor on the running machine (as determined by Etc.nprocessors).
• #97 - Syntax Tree now handles the case where DidYouMean is not available for whatever reason, as well as handles the newer detailed_message API for errors.

2.7.1 - 2022-05-25

Added

• #92 - (Internal) Drastically increase test coverage, including many more tests for the language server and the CLI.

Changed

• #87 - Don't convert quotes on strings if it would result in more escapes.
• #91 - Always use [] with array patterns. There are just too many edge cases where you have to use them anyway. This simplifies the look and makes it more consistent.
• #92 - Remodel the currently shipped plugins such that they're modifying an options hash instead of overriding methods. This should make it easier for other plugins to reference the already loaded plugins, e.g., the RBS plugin referencing the quotes.
• #92 - Fix up the language server inlay hints to continue walking the tree once a pattern is found. This should increase useability.

2.7.0 - 2022-05-19

Added

• #88 - Provide a SyntaxTree::BasicVisitor that has no visit methods implemented.

Changed

• #90 - Provide better formatting for SyntaxTree::AryPtn when its nested inside a SyntaxTree::RAssign.

2.6.0 - 2022-05-16

Added

• #74 - Add Rake test to run check and format commands.
• #83 - Add a trailing commas plugin.
• #84 - Handle lambda block-local variables.

Changed

• #85 - Better handle trailing operators on command calls.

2.5.0 - 2022-05-13

Added

• #79 - Support an optional maxwidth second argument to SyntaxTree.format.

Changed

• #77 - Correct the pattern for checking if a dynamic symbol can be converted into a label as a hash key.
• #72 - Disallow conditionals with not without parentheses in the predicate from turning into a ternary.

2.4.1 - 2022-05-10

• #73 - Fix nested hash patterns from accidentally adding a then to their output.

2.4.0 - 2022-05-07

Added

• #65 - Add a rubocop config at config/rubocop.yml that we can ship with the gem so folks can inherit from it to get their styling correct.
• #65 - Improve hash pattern formatting by a lot - multiple lines are now not so ugly.
• #62 - Add options as a method on SyntaxTree::RegexpLiteral, add it to pattern matching, and describe it using the SyntaxTree::Visitor::FieldVisitor class.
• #69 - The construct_keys option has been added to every SyntaxTree::Node descendant. This allows building a pattern match expression that can be used later. It is meant as a reflection API, not necessarily something that should be eval'd.
• #69 - You can now call stree json to get a JSON representation of your syntax tree.
• #69 - You can now call stree match to get a Ruby pattern matching expression to match against the given input.

Changed

• #69 - Fixed a long-standing bug with pretty-print where if certain things were required in different orders you could end up with a bug in PP when calling pretty-print with a confusing error referring to inspect keys.
• #69 - SyntaxTree.read can now handle an empty file.

2.3.1 - 2022-04-22

Changed

• SyntaxTree::If nodes inside of SyntaxTree::Command arguments should include a space before if they are flat.

2.3.0 - 2022-04-22

Added

• #52 - SyntaxTree::Formatter.format for formatting an already parsed node.
• #56 - if and unless can now be transformed into ternaries if they're simple enough.
• #56 - Nicely format call chains by one indentation.
• #56 - Handle trailing operators in call chains when they are necessary because of comments.
• #56 - Add some specialized formatting for Sorbet sig blocks to make them appear nicer.

Changed

• #53 - Optional keyword arguments on method declarations have a value of nil now instead of false. This makes it easier to use the visitor.
• #54 - Flow control operators can now skip parentheses for simple, individual arguments. e.g., break(1) becomes break 1.
• #54 - Don't allow modifier conditionals to modify ternaries.
• #55 - Skip parentheses and brackets on arrays for flow control operators. e.g., break([1, 2, 3]) becomes break 1, 2, 3.
• #56 - Don't add parentheses to method calls if you don't need them.
• #56 - Format comments on empty parameter sets. e.g., def foo # bar should keeps its comment.
• #56 - %s[] symbols on assignments should not indent to the next line.
• #56 - Empty hash and array literals with comments inside of them should be formatted correctly.

2.2.0 - 2022-04-19

Added

• #51 - SyntaxTree::Location nodes now have pattern matching.
• #51 - SyntaxTree::Heredoc now have a dedent field that indicates the number of spaces to strip from the beginning of the string content.

Changed

• #51 - SyntaxTree::HshPtn will now add a then if you use a bare ** and SyntaxTree::AryPtn will do the same for a bare * on the end.
• #51 - SyntaxTree::MLHSParen now has a comma field in case a trailing comma has been added to a parenthesis destructuring, as in ((foo,)).
• #51 - SyntaxTree::FndPtn has much improved parsing now.

2.1.1 - 2022-04-16

Changed

• #45 - Fix parsing expressions like foo.instance_exec(&T.must(block)), where there are two args_add_block calls with a single &. Previously it was associating the & with the wrong block.
• #47 - Handle expressions like not().
• #48 - Handle special call syntax with :: operator.
• #49 - Handle expressions like case foo; in {}; end.
• #50 - Parsing expressions like case foo; in **nil; end.

2.1.0 - 2022-04-12

Added

• The SyntaxTree::Visitor class now implements the visitor pattern for Ruby nodes.
• The SyntaxTree::Visitor.visit_method(name) method.
• Support for Ruby 2.7.
• Support for comments on rescue and else keywords.
• SyntaxTree::Location now additionally has start_column and end_column.
• The CLI now accepts content over STDIN for the ast, check, debug, doc, format, and write commands.

Removed

• The missing hash value inlay hints have been removed.

2.0.1 - 2022-03-31

Changed

• Move the SyntaxTree.register_handler method to the correct location.

2.0.0 - 2022-03-31

Added

• The new SyntaxTree.register_handler hook for plugins.
• The new --plugins= option on the CLI.

Changed

• Changed SyntaxTree from being a class to being a module. The parser functionality is moved into SyntaxTree::Parser.
• There is now a parent class for all of the nodes named SyntaxTree::Node.
• The Implicits class has been renamed to InlayHints to match the new LSP spec.

Removed

• The disassembly code action has been removed to limit the scope of this project overall.

1.2.0 - 2022-01-09

Added

• Support for Ruby 3.1 syntax, including: blocks without names, hash keys without values, endless methods without parentheses, and new argument forwarding.
• Support for pinned expressions and variables within pattern matching.
• Support endless ranges as the final argument to a when clause.

1.1.1 - 2021-12-09

Added

• #7 Better formatting for hashes and arrays that are values in hashes.
• #9 Special handling for RSpec matchers when nesting CommandCall nodes.
• #10 Force the maintaining of the modifier forms of conditionals and loops if the statement includes an assignment. Also, for the maintaining of the block form of conditionals and loops if the predicate includes an assignment.

1.1.0 - 2021-12-08

Added

• Better handling for formatting files with errors.
• Colorize the output snippet using IRB.

1.0.0 - 2021-12-08

Added

• The ability to "check" formatting by formatting the output of the first format.
• Comments can now be attached to the case keyword.
• Remove escaped forward slashes from regular expression literals when converting to %r.
• Allow arrays of CHAR nodes to be converted to QWords under certain conditions.
• Allow HashLiteral opening braces to have trailing comments.
• Add parentheses if Yield breaks onto multiple lines.
• Ensure all nodes that could have heredocs nested know about their end lines.
• Ensure comments on assignment after the = before the value keep their place.
• Trailing comments on parameters with no parentheses now do not force a break.
• Allow ArrayLiteral opening brackets to have trailing comments.
• Allow different line suffix nodes to have different priorities.
• Better support for encoding by properly reading encoding magic comments.
• Support singleton single-line method definitions.
• Support stree-ignore comments to ignore formatting nodes.
• Add special formatting for arrays of VarRef nodes whose sum width is greater than 2 * the maximum width.
• Better output formatting for the CLI.

Changed

• Force a break if a block is attached to a Command or CommandCall node.
• Don't indent CommandCall arguments if they don't fit aligned.
• Force a break in Call nodes if there are comments on the receiver.
• Do not change block bounds if inside of a Command or CommandCall node.
• Handle empty parentheses inside method calls.
• Skip indentation for special array literals on assignment nodes.
• Ensure a final breakable is inserted when converting an ArrayLiteral to a QSymbols.
• Fix up the doc_width calculation for CommandCall nodes.
• Ensure parameters inside a lambda literal when there are no parentheses are grouped.
• Ensure when converting an ArrayLiteral to a QWords that the strings do not contain [.
• Stop looking for parent Command or CommandCall nodes in blocks once you hit Statements.
• Ensure nested Lambda nodes get their correct bounds.
• Ensure we do not change block bounds within control flow constructs.
• Ensure parentheses are added around keywords changing to their modifier forms.
• Allow conditionals to take modifier form if they are using the then keyword with a VoidStmt.
• UntilMod and WhileMod nodes that wrap a Begin should be forced into their modifier forms.
• Ensure For loops keep their trailing commas.
• Replicate content for __END__ keyword exactly.
• Keep block If, Unless, While, and Until forms if there is an assignment in the predicate.
• Force using braces if the block is within the predicate of a conditional or loop.
• Allow for the possibility that CommandCall nodes might not have arguments.
• Explicitly handle ?" so that it formats properly.
• Check that a block is within the predicate in a more relaxed way.
• Ensure the Return breaks with brackets and not parentheses.
• Ensure trailing comments on parameter declarations are consistent.
• Make Command and CommandCall aware that their arguments could exceed their normal expected bounds because of heredocs.
• Only unescape forward slashes in regular expressions if converting from slash bounds to %r bounds.
• Allow When nodes to grab trailing comments away from their statements lists.
• Allow flip-flop operators to be formatted correctly within IfMod and UnlessMod nodes.
• Allow IfMod and UnlessMod to know about heredocs moving their bounds.
• Properly handle breaking parameters when there are no parentheses.
• Properly handle trailing operators in call chains with attached comments.
• Force using braces if the block is within the predicate of a ternary.
• Properly handle trailing comments after a then operator on a When or In clause.
• Ensure nested HshPtn nodes use braces.
• Force using braces if the block is within a Binary within the predicate of a loop or conditional.
• Make sure StringLiteral and StringEmbExpr know that they can be extended by heredocs.
• Ensure Int nodes with preceding unary + get formatted properly.
• Properly handle byte-order mark column offsets at the beginnings of files.
• Ensure Words, Symbols, QWords, and QSymbols properly format when their contents contain brackets.
• Ensure ternaries being broken out into if...else...end get wrapped in parentheses if necessary.

Removed

• The AccessCtrl node in favor of just formatting correctly when you hit a Statements node.
• The MethodAddArg node is removed in favor of an optional arguments field on Call and FCall.

0.1.0 - 2021-11-16

Added

• 🎉 Initial release! 🎉