2026.05.15¶
~literalizer.literalizenow accepts an opt-inbound_refsmapping. Unlikeref_values(which only informs a ref’s type and leaves it as a free external identifier), each name inbound_refsadditionally has a binding emitted for it before its first use, so a singleliteralize(..., bound_refs=..., wrap_in_file=True)call produces a complete, valid file with per-language declaration sequencing (Nix nestedlet, the Fortran rule that specification statements precede executable statements, and so on). Binding emission only happens withwrap_in_file=Trueand a~literalizer.NewVariableor~literalizer.ExistingVariablevariable_form; otherwisebound_refsdegrades to type information only, exactly likeref_values. The default behavior is unchanged. Withoutbound_refs, a$refis a free external identifier. The ref golden-file harness now drives every case through oneliteralizecall, retiring its regex-based stub-stitching helpers. See #2294.~literalizer.Gogains theRECORDheterogeneous_strategy(already on~literalizer.Rust). Each record-shaped dict (non-empty, string-keyed) becomes a generatedtype RecordN struct { ... }declared in the preamble plus a matchingRecordN{Field: value, ...}literal, so a dict whose values mix scalars and containers is representable instead of raising. Field names are exported (PascalCase) and the struct-name prefix is configurable via the newrecord_struct_name_prefixconstructor parameter. See #2297.~literalizer.Scalagains theRECORDheterogeneous_strategy(already on~literalizer.Rustand~literalizer.Go). Each record-shaped dict (non-empty, string-keyed) becomes a generatedcase class RecordN(field: Type, ...)declared in the enclosingobjectplus a matchingRecordN(field = value, ...)literal, so a dict whose values mix scalars and containers is representable instead of raising. Field names are the dict keys verbatim and thecase class-name prefix is configurable via the newrecord_struct_name_prefixconstructor parameter. See #2299.~literalizer.Rustaccepts arecord_shape_namesconstructor parameter, a mapping from each record’s key-set (frozenset[str]) to a user-chosenstructname, so theRECORDheterogeneous strategy can emitstruct Entry { ... }instead of the auto-generatedRecord0,Record1, … names. Shape names that are not PascalCase Rust identifiers, that collide withheterogeneous_value_enum_name, that duplicate another mapped name, or that match a Rust reserved keyword raise the new~literalizer.exceptions.InvalidRecordNameError. The existingrecord_struct_name_prefixis validated the same way. See #2236.~literalizer.Fortrannow offersVersionFormats.V2003alongsideVersionFormats.V2008(the default). The 2003 target defines theint64kind viaselected_int_kind(18)instead of importing it from the intrinsiciso_fortran_envmodule (whoseint64constant is a Fortran 2008 addition), so generated code is otherwise identical and the_int64literal suffix is unchanged. The golden harness emits a parallel@v2003/@v2008fixture set and CI lints each with the matchinggfortran -std=f2003/-std=f2008flag. See #1931.The integration golden-file harness now accepts
input.tomlnext to the existinginput.yamlfor cases whose input contains a value YAML 1.2 cannot natively express (currentlydatetime.time). Newtime_list,time_dict, andtimes_heterogeneous_with_datesgolden cases give every supported language end-to-end coverage ofdatetime.timerendering, which surfaced and fixed several preexisting bugs in time emission: tagged-union languages (~literalizer.Elm,~literalizer.OCaml,~literalizer.Roc,~literalizer.Gleam,~literalizer.Haskell,~literalizer.Zig,~literalizer.Sml,~literalizer.Ada,~literalizer.Fortran,~literalizer.ObjectiveC) now wrap a fallback ISO 8601 time inside the value-type constructor the same way they wrap other scalars; collection type inference (C++, C#, Java, Kotlin, Scala, Dart, Go, Visual Basic, and others) now knows aboutdatetime.timeand narrows homogeneous collections to the correct element type; and~literalizer.FSharpfully qualifiesSystem.TimeOnlyso the rendered output no longer emits anopen Systemline beforemodule. See #2230.~literalizer.Haskellnow acceptsvariable_formon~literalizer.literalize_call, emitting the inference-style bindingmy_data = make_widget (42)without aname :: Typeannotation (the call’s return type is not known to the renderer).wrap_in_file=TrueHaskell scaffolds emit a{-# OPTIONS_GHC -Wno-missing-signatures #-}pragma so the inferred binding compiles under-Wall -Werror. See #2244.datetime.timeis now a first-classScalarvalue. Languages with a native time-only type emit native literals (Pythondatetime.time(...), TOML unquotedHH:MM:SS, .NETnew TimeOnly(...)/TimeOnly(...)/New TimeOnly(...)for~literalizer.CSharp,~literalizer.FSharp,~literalizer.VisualBasic, andLocalTime.of(...)for~literalizer.Java,~literalizer.Kotlin,~literalizer.Scala,~literalizer.Groovy); other languages fall back to the existing ISO 8601 quoted-string form. TOMLtimeinputs now round-trip through~literalizer.Tomlas native time literals instead of being re-emitted as quoted ISO 8601 strings. Typed collection openers in languages such as~literalizer.CSharp,~literalizer.Java,~literalizer.Kotlin, and~literalizer.Scalanow narrow uniform time-only sequences to the language’s time type (e.g.TimeOnly[]/Array<LocalTime>) instead of falling back to a genericObject/Anyopener.Golden files for languages whose compiler version is pinned (Elixir, Erlang, Gleam, Kotlin, Odin, Zig) now carry the version in the filename:
{stem}@{version}{extension}(e.g.Odin@dev-2026-04.odin). Every fixture is explicitly tied to the compiler version it was generated against. Eachlint.ymljob sets a job-scoped environment variable once and feeds it to both the install action andpython -m tests.integration.list_fixtures; the test code auto-discovers the same version from sibling filenames so no separate registry file is needed.~literalizer.Fortran’slanguage_versiondefault is nowFortran.VersionFormats.V2008(wasV2003) so it matches the Fortran 2008 features the generator actually emits (e.g.int64fromiso_fortran_env).V2003has been removed fromFortran.VersionFormats.lint-fastCI job now syntax-checks and runs the Python fixtures undertests/integration/cases, matching the per-language gate already in place for Bash, Ruby, JavaScript, and other fixture languages. See #1921.lint-odinCI job now usesodin run .again to catch runtime errors thatodin build .cannot detect (e.g. nil-proc calls). Thelaytan/setup-odinaction is pinned todev-2026-04, the last Odin release whereodin run .did not segfault on these fixtures;release: latest(dev-2026-05) crashes at runtime, and the compiler itself segfaults on some fixtures underodin build .. See #1745.