Changelog

2026.08.30.1

No documented changes.

2026.08.30

No documented changes.

2026.08.29

• Refuse empty input rather than answering differently per format. The same empty input produced a JSON parse error, None for YAML and an empty dict for TOML.

  Strip a leading byte order mark from input. It is a decoding artifact rather than data, and reached the parsers as content.

  Add --input-file to read input from a path instead of standard input.

  Refuse a --modifier given more than once, which previously collapsed into the modifier set with no feedback.

  Warn when --include-preamble is given for output that has no preamble.

  Report the literalizer version alongside the literalize version, since the library decides what the output looks like.

  Keep the originating library exception attached when reporting an error, rather than discarding it with from None.

  Say in --help that language-option choices are the union across every language, since --help renders before --language is known.

• Reject formatting options that produce malformed output. --indent no longer accepts an empty string or one holding a line break, --ref-key no longer accepts an empty or whitespace-only marker, and --pre-indent-level is now bounded because each level repeats the indent on every output line.

• Reject options that the selected mode ignores. --call-function, --call-params and --per-element in literal mode, and --pre-indent-level and --include-delimiters in call mode, previously had no effect and produced output that did not reflect the request. --no-new-variable without --variable-name was a silent no-op and is now refused too.

• Report a --call-params value that names no parameter as such. An empty, whitespace-only or comma-only value previously reached the arity check, which reported “Expected 0 parameters but got N values”.

• Report every literalizer error as a clean CLI message. The CLI matched two hand-maintained tuples of exception classes, so the 32 LiteralizerError subclasses missing from them escaped as Python tracebacks. Both sites now match the LiteralizerError base class.

• Sign the macOS binary with a Developer ID certificate and notarize it, so Gatekeeper no longer blocks it when it is downloaded in a browser. The xattr -d com.apple.quarantine workaround is no longer needed.

2026.08.16.1

No documented changes.

2026.08.16

• Add manual Sphinx linkcheck and spelling pre-commit hooks to complete the lint tooling alignment with literalizer.

• Bump literalizer to 2026.7.24.1. The CLI now offers C++14, C++17, and C++20 through --language-version and exposes generated heterogeneous-value carrier names through --heterogeneous-value-variant-name. The release also adopts upstream fixes for RECORD shape inference, null-byte strings, empty containers, wide integers, and variable-name validation; invalid or reserved new variable names are surfaced as clean CLI errors.

• Expose Python annotation evaluation and union format options.

• Add --record-struct-name-prefix for naming the structs/records that the record --heterogeneous-strategy generates (e.g. Widget0, Widget1 instead of Record0, Record1). The record strategy is now available for many more languages in this release (C, C#, C++, Crystal, D, Go, Java, Kotlin, Nim, Odin, Python, Rust, Scala, Swift, V, Zig). The strategy and the new option are surfaced automatically per language exactly like every other language-specific option.

• An invalid --record-struct-name-prefix (not a PascalCase identifier for the target language) now surfaces the upstream InvalidRecordNameError as a clean CLI error rather than a traceback.

• Bump literalizer to 2026.5.17.

• Bump literalizer to 2026.8.2. Native multiline string formats are now available through --string-format multiline for supported languages, and --multiline-raw-string-delimiter-base configures fallback delimiters for C++ raw strings. Invalid C++ delimiter bases are surfaced as clean CLI errors.

• --heterogeneous-strategy tuple is now offered for the languages that gained the upstream TUPLE strategy (C++, Kotlin, Rust, Scala, TypeScript). A tuple arity that has no native fixed-size tuple in the target language (e.g. a 4+-element heterogeneous array in Kotlin) now surfaces the new upstream TupleArityNotRepresentableError as a clean CLI error rather than a traceback.

• literalizer removed DottedCallStubNotSupportedError and FreeFunctionCallNotSupportedError because the context-aware call_transform made them unreachable. They are no longer referenced. Languages whose declaration template previously only wrapped literal values (Bash, Objective-C, Tcl, and others) now bind a call result through their idiomatic call-binding form, so --variable-name in --mode call works for them too.

2026.05.14.1

• Bump literalizer to 2026.5.14.1.

• YAML inputs with non-string dict keys (integers, dates, booleans) now flow through to the target language’s value-formatting path instead of being silently stringified. Languages that can represent the key natively (Python, Ruby, Clojure, Lua, Bash, and others) produce the corresponding literal. Languages whose dict syntax requires string keys or a homogeneous typed map surface the new upstream UnrepresentableInputError as a clean CLI error.

2026.05.14

• Bump literalizer to 2026.5.14.

• --variable-name (with --new-variable / --no-new-variable and --modifier) now applies in --mode call as well, wrapping the rendered call in the language’s idiomatic per-language variable binding (e.g. let user = createUser(...), const user = createUser(...), user = create_user(...)). Mutability and inference are picked up from --declaration-style and --modifier exactly as in literal mode. Languages whose declaration template wraps or transforms the right-hand side in a way only valid for literal values (e.g. Bash command substitution, Objective-C boxing, tagged-enum heterogeneous-strategy languages) surface the upstream UnsupportedCallShapeError as a clean CLI error, as does the combination with --per-element (which has no per-element name vector).

2026.05.13

• Bump literalizer to 2026.5.13.1. The new release re-exposes the supports_* class attributes for empty_dict_key, call_style, and the five default_*_type options, restoring a type-safe probe for runtime-dispatched constructor kwargs (cf. upstream issue #2147).

• Replace --line-ending with --statement-terminator-style. The upstream LineEndings enum was removed; StatementTerminatorStyles (semicolon, none) is its successor.

• Add --call-style for picking between per-language call shapes (e.g. curried for Haskell / OCaml / F# / SML / Elm, named for Visual Basic).

• Add --numeric-style for languages that support multiple numeric rendering styles (e.g. overloaded vs. explicit).

• Add --language-version for selecting the target language version (each language exposes a VersionFormats enum).

• Add --module-name for languages whose --wrap-in-file form introduces a named scope (C, C++, D, Erlang, Fortran, F#, Java, Objective-C, Occam, SystemVerilog). Previously a module_name argument to literalize itself, now a per-language constructor argument.

• Add --ref-key for picking a marker key other than $ref for variable-reference mappings in the input data.

• Surface the new typed literalizer exceptions as clean CLI errors rather than tracebacks: UnsupportedCallShapeError, VariableNameNotSupportedError, WrapInFileWithoutVariableNotSupportedError, WrapCombinedInFileNotSupportedError, DottedCallTargetNotSupportedError, DottedCallStubNotSupportedError, FreeFunctionCallNotSupportedError, CallArgNotSupportedError, HeterogeneousScalarCollectionError, UnrepresentableSpecialFloatError.

• --variable-type-hints auto is now --variable-type-hints never (upstream rename), with a new safe option that annotates only when the language’s own inference would widen the variable to a permissive type (e.g. unknown[] for an empty TypeScript array).

2026.04.30

• Bump literalizer to 2026.4.29 (adds Roc, Wren, Mojo, V, Ada, Nim, Tcl, Scheme, PureScript, OCaml, SystemVerilog, COBOL, Fortran, Dart, Dhall, Elixir, Elm, and PowerShell to literalize_call support).

• Add --ref-case for emitting $ref markers in input data as bare identifiers re-cased to snake, camel, pascal, upper_snake, or kebab.

• Bump literalizer to 2026.4.21.4.

• Add --heterogeneous-strategy to pick between per-language strategies for collections with mixed scalar types (e.g. Rust’s tagged_enum, which emits a generated tagged enum preamble and wraps each value at the call site).

• Surface literalize_call errors as clean CLI messages rather than tracebacks: parameter-count mismatches, languages with no call syntax (YAML, TOML, JSON5, Norg), and languages whose call rendering is not yet implemented all now exit with a descriptive Error: line.

• Add --modifier (repeatable) for declaration modifiers on new variables in languages that support them (Java, C#, C++).

• Remove --error-on-coercion: literalizer now always errors on heterogeneous data that cannot be represented in the target language.

• Add --mode call for converting data into function call expressions, with --call-function, --call-params, and --per-element options.

2026.04.06

2026.03.29

• Bump literalizer to 2026.03.26.1.

• Replace --line-prefix CLI option with --pre-indent-level.

2026.03.25

2026.03.23.7

2026.03.23.6

2026.03.23.5

2026.03.23.4

2026.03.23.3

2026.03.23.2

2026.03.23.1

2026.03.23