Skip to content

Latest commit

 

History

History
93 lines (63 loc) · 5.01 KB

CHANGELOG.md

File metadata and controls

93 lines (63 loc) · 5.01 KB
Raw

Change Log

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

[2.0.0] - 2024-11-06

New

  • GH-50 - term rewrite blocks, including adding
    • reserved keyword leafs term ("term"), signature ("signature"), rewrite ("rewrite"), dot ("."), and rewrite_arrow ("==>")
    • named nodes rewrite_case, rewrite_term, rewrite_type, and rewrite_block. The fourth is an entire rewrite block beginning with @rewrite, and it, inter alia, includes at least one of the other three as a child node
  • GH-85 - a built-in hash is a single unit, no longer *PATH_SEGMENT . REGULAR_ID
  • GH-86 - add force, in which x() is parsed as (force (regular_identifier) (unit))
  • GH-90 - liberal recognition of _layout_semicolon at ends of lines with equal indentation levels, which helps with parsing in many places
  • GH-91 - infix symops can have namespace/path prefix

Fixed

  • GH-80 - parsing folds of more than three hyphens
  • GH-84 - Symbols in paths now work
  • GH-85 - built-ins are now just built_in_hash without any child nodes
  • GH-92 - test watch expressions can only have test and test.io as prefix for >
  • GH-95 - doc blocks can only appear two ways (syntax sugar before a top-level binding and in similar places as a literal)
  • GH-96 - enable built-in hashes
  • GH-97 - separate a term binding from a top-level binding (latter has the doc block syntax sugar)
  • GH-98 - +4 is correctly tokenized as int rather than symboly_id . nat
  • #?? - pattern matching guards now are part of a layouted block
  • GH-103 - match scrutinee now correctly categorized as _term rather than _expression
  • GH-105 - two-char symbols ending in & or | now parse as two-char symbols

Changed

  • wordy_id now called regular_identifier in most places
  • namespace now includes path and optional regular_identifier/operator child nodes
  • wordy_id has been changed to ability_name and type_argument in ability declarations
  • wordy_id has been changed to constructor_name in ability constructors
  • colon has been added as a named token in ability constructors
  • wordy_id has been changed to field_name in record fields
  • guard has been renamed guarded_block
  • the actual guard of a guarded block is its own node now, called guard, and can be the parent of various nodes
  • ability_pattern has been renamed effect_pattern to be in line with UCM codebase terminology
  • effect_pattern now contains either a (new) effect_pure or effect_bind. The difference is as that between { x } and { x -> y }, respectively
  • constructor_or_variable_pattern has been bifurcated into ctor and var_or_as because they have subtly different rules than anticipated previously so cannot be handled together
  • the list infix patterns have been flattened, so for example init_last_tail_pattern ( left side . right side ) is now left side . snoc . right side. The sructure can be deduced, while now the infix operator is a named node. Another example is head_tail_list_pattern (left . right) is now left cons right.
  • a prefix operator (which is surrounded by parentheses) is now called prefix_operator instead of just operator
  • the introduction of named nodes for [, ], (, and ): open_bracket, close_bracket, open_parens, close_parens
  • the introduction of named node comma in tuple patterns

[1.1.4] - 2024-04-25

Fixed

  • #73 - Removed bounds checks (atof) for int, nat, float and now use JS rather than C to parse

[1.1.3] - 2024-04-24

Fixed

  • #71 - Hashes now only validate for proper base32hex (i.e., fail if a letter after 'v' appears in the hash)
  • #67 - fprintf conditionally included in scanner so WASM builds succeed

[1.1.1] - 2024-04-13

Fixed

  • #57 - strncat instead of strcat to support WASM

[1.1.0] - 2024-02-02

Added

  • #54 - Unison now defaults to "unique" for abilities and types, so we allow a tacit unique for them in the grammar
  • #52 - Unison allows hex literals for natural numbers. This was not supported until Felipe Guerreiro filed a PR, and I got off my lazy kiester to test and merge it in.

Fixed

  • #55 - There were two sets of unique and structural keywords in the grammar, so they've been consolidated now. This might require updates to anything relying on the output of "unique_kw" and "structural_kw" for types, as they now output to "unique" and "structural" like abilities always have. Thank you to @zetashift for reporting this!

[1.0.2] - 2023-07-28

Fixes

  • #39 - term declarations now take both path-prefixed wordy ids as well as path-prefixed symboly ids that are in parentheses.

[1.0.1] - 2023-07-26

Fixes

  • Previous version allowed documentation blocks anywhere. They are only supposed to be allowed right before term definitions (as "anonymous documentation blocks") or as a type of expression.

[1.0.0] - 2023-06-28

All Unison entities parsed, but doc blocks are only parsed as a doc block without breaking them down further into sections, etc.