| Commit message (Collapse) | Author | Files | Lines |
|---|
|
- Fixed some minor bugs.
- Added the updated code for the parser (forgot to add it to a previous commit).
|
|
|
|
with type-inference & Type/Show needs some pending corrections).
|
|
on Void to achieve "recursion".
- Introduced ExQ types into the type-system (still pending work on inference).
|
|
- Corrected some code that still involved the old names for the list macros.
- Corrected some code in the pattern-matcher analyser to it fails properly when encountering invalid pattern-syntax.
|
|
each other when saving the class files in case-insensitive file-systems (https://github.com/LuxLang/lux/issues/8). The names of certain definitions were changed slightly to avoid clashes and the compiler throws an error if the names end up clashing prior to saving the .class file.
|
|
- Added some minor compiler optimizations.
|
|
- Implemented hygienic macros by adding global symbol resolution inside the ` macro.
|
|
- Improved the "struct" macro so that (once again) it doesn't ask for the prefixes of the members.
- Added tests for the lux.reader namespace.
|
|
- Removed the (unnecessary) lux/control/dict & lux/control/stack modules.
- The "Meta" type is now a record instead of a variant.
|
|
- Removed the (unnecessary) lux/data/error module and moved it's structures to lux/data/either.
- Implemented the \slots destructurer for records.
- Implemented quicksort for lists as the "sort" function in lux/data/list.
- Added tags for the Cursor type.
|
|
- UnivQ no longer stores the environment as key-val pairs with Text names, but instead stores it as type-lists with variables accessed via an index through a (updated) BoundT.
- UnivQ no longer stores the name of the type-fun, not the name of the type-arg.
|
|
|
|
|
|
to too many corner cases, I decided the abandon the path towards a more mathematical implementation of tuples & variants.
|
|
|
|
|
|
|
|
- As an unexpected bonus, the compiler has become 2.5x faster.
- Fixed some minor bugs.
- Tag declarations now include associated types.
- Tag declarations info is now stored twice (one from the perspective of tags, the other from the perspective of types).
- Changed the named of the "types" member of the Compiler type, to "type-vars" to avoid collision with the "types" member of the Module type.
|
|
|
|
|
|
- some adjustments to the new format for types.
|
|
- Bug fixes and adjustments.
|
|
- Added several bug fixes
- Changed "Reader" to "Source" and "HostState" to "Host" in lux.lux
- Set up indexing of records via tags.
- Added lux.analyser.record namespace.
- Removed some (now) unnecessary code for working with records.
- Added the license (can't believe I missed it for so long.)
|
|
- Fixed some bugs.
- Now pattern-matching on variants works with indices, rather than text tags.
|
|
- Fixed a bug regarding type coercion (type-checking was ocurring unnecessarily).
- Fixed another bug regarding Local/Global variables.
|
|
|
|
avoid the problem of error ocurring "nowhere" (at ["" -1 -1]).
|
|
prohibits them).
|
|
- Created the lux/meta/ast module.
|
|
on definitions).
- The "open" and "using" macros now work recursively on records.
|
|
- Value definitions inside the compiler data now hold the value itself.
- Fixed a few bugs.
|
|
|
|
- Refactored the standard library a bit.
- Implemented the "loop" macro.
- Added the expected type of expressions as a field in the compiler state.
- Added syntactic sugar for using tuples with variants, in order to minimize the usage of brackets to delimit the contents of data-structures.
- Fixed a bug wherein "macro-expand" was behaving like "macro-expand-all", and added a separate implementation for "macro-expand-all".
- Fixed a few bugs.
|
|
- Introduced 2 new modules: lux/data/tuple & lux/codata/function
- Now doing safe reading of files.
- Took the "let", "lambda" & "def" macros to their ultimate form.
- Added some macros for doing better JVM interop.
- Fixed a bug when compiling comparisons for doubles.
- Changed the order in which arguments are compiled for all arithmetic operations, as the order is reversed (from the conventional order) in the JVM bytecode.
|
|
- Fixed several bugs with host (JVM) interop.
- Now packaging everything in a .jar file ("program.jar").
|
|
- Completed lux/host/jvm
- Modified (slightly) the syntax used in several host (JVM) special forms.
- The "defsyntax" macro now binds all of the arguments it receives inside a variable named "tokens".
|
|
(universal-quantification / AllT) was not being taken into account properly when destructuring.
- Fixed a compiler error wherein the types of definitions didn't generate (correctly) the structures necessary for storage inside the class _meta(data) field.
- Improved both the "open" and "import" macros with extra features.
|
|
- "input" has been renamed as "source" and "output" has been renamed as "target".
|
|
- Implemented module-caching to avoid the waiting too much during program compilation.
|
|
- Now storing information about definitions & imports inside the .class files.
|
|
|
|
|
|
- The version of the compiler is now stored as a field in the compiled definitions.
|
|
- The command-line params argument in jvm_program is now transformed from a String array into (List Text).
|
|
|
|
- Replace most instances of "=" with ".equals".
- Added an optimization to lux.type/type= that drastically speeds-up type comparisons.
|
|
type-functions which are most-likely the same, due to their name (remembering that when you define types using deftype, the type-function's name will correspond to the def's).
- Gave empty environments to top-level type-functions, instead of leaving them with unset environments.
|
|
- Temporarily reverted back to forward apply-analysis.
- Fixed an error in lux.base/show-ast.
- Reader now only returns a tuple instead of a full-blown #Meta variant.
- Reader now doesn't cut the strings that it reads. Instead, the "cursor" just moves around, indicating where to read.
- Inlined some calculations that previously relied on try-all%.
|
|
|
