lux - fork of https://github.com/LuxLang/lux for my bookshelf

	Commit message (Collapse)	Author	Files	Lines
2016-06-13	- Implemented a procedure for loading JVM classes.	Eduardo Julian	2	-0/+24

2016-06-13	- Removed all the unnecessary stack-frame descriptors.	Eduardo Julian	1	-15/+0

2016-06-13	- Now avoiding unnecessary pops of the pattern-matching stack/cursor.	Eduardo Julian	1	-3/+11

2016-06-13	- Finished the test-sharing code.	Eduardo Julian	6	-398/+401

2016-06-06	- [WIP] sharing tests during pattern-matching.	Eduardo Julian	4	-177/+402

2016-05-22	- Fixed that bug that was causing method definitions to be set as final and ↵	Eduardo Julian	2	-6/+10
	with strict-fp math.
2016-05-22	- Added Tail-Recursion Optimization.	Eduardo Julian	5	-104/+177

2016-05-21	- Merged _lux_alias into _lux_import.	Eduardo Julian	3	-30/+26

2016-05-21	- Now it's possible to pattern-match against variants using numeric indices.	Eduardo Julian	3	-20/+15

2016-05-21	- Now folding top-level function-classes into their owner def-classes.	Eduardo Julian	4	-100/+209

2016-05-21	- Improved the optimization of self-calls.	Eduardo Julian	1	-4/+9

2016-05-21	- Fixed a bug in which it was impossible to pattern-match against ↵	Eduardo Julian	7	-67/+169
	existentially-qualified types. - Improved error-reporting. - When loading a class post-compilation, the ClassLoader kept referring to the previous dummy version used during analysis, which meant the real class, with it's code, couldn't be used at compile time. Fixed this (with a hack, sadly...). - Fixed a bug in which using JVM type-vars with top-bounds different from java.lang.Object was not getting acknowledged by the compiler, and resulted in incorrect signatures for methods.
2016-05-20	- Fixed a bug when parsing the names of arguments in method definitions, ↵	Eduardo Julian	2	-2/+6
	wherein the names couldn't be full-blown (unprefixed) idents.
2016-05-14	- Now storing the type-variables created during type-checking for the sake ↵	Eduardo Julian	4	-48/+93
	of reference later on in definitions.
2016-05-13	- Optimized pattern-matching on the analyser-phase.	Eduardo Julian	4	-177/+135
	- Analysis of statements now ensures they are top-level forms, by making sure there is no expected type when they are analyzed.
2016-05-13	- The expected-type in the compiler-state is now a (Maybe Type), instead of ↵	Eduardo Julian	5	-165/+184
	a Type. - Simplified and optimized the pattern-matching done during the analysis phase. - Fixed the bug that caused the JVM garbage-collector to go crazy after running the compiler.
2016-05-13	- Fixed a bug that allowed pattern-matching using unknown tags to proceed.	Eduardo Julian	5	-27/+72
	- Streamlined invocation of functions when the args-list >= the arity. The impl method gets called immediately, instead of passing first by the apply method.
2016-05-12	- Now with better localization for reader, lexer & parser errors.	Eduardo Julian	6	-24/+24

2016-05-11	- Now generating variants of the "apply" method of lux/Function up to 8 arities.	Eduardo Julian	7	-112/+70

2016-05-11	- Now fusing impl and apply1 for 1-arity functions.	Eduardo Julian	1	-80/+90
	- Now avoiding to "reset" functions at stage 0, when called with an apply-N where N = arity.
2016-05-11	- lux/Function is now an abstract class with 4 versions of apply, to improve ↵	Eduardo Julian	6	-69/+176
	performance when calling functions.
2016-05-10	- Added an optimization to fold directly-nested functions and fuse them into ↵	Eduardo Julian	13	-296/+640
	multi-argument functions.
2016-05-08	- Removed _jvm_class, _jvm_interface and _jvm_anon-class from the list of ↵	Eduardo Julian	10	-675/+732
	special forms.
2016-05-03	- Made a small adjustment when packaging the .jars of programs, to avoid ↵	Eduardo Julian	1	-1/+4
	packaging .jars that are necessary for the compiler itself, but not for the Lux program that was compiled.
2016-05-03	- Added multi-platform support for arrays.	Eduardo Julian	2	-0/+83

2016-05-03	- Now, showing the right error when trying to lex an ident with an ↵	Eduardo Julian	5	-47/+94
	invalid/unknown prefix.
2016-05-03	- Renamed #DataT to #HostT.	Eduardo Julian	12	-117/+126
	- Renamed _lux_host to _lux_proc.
2016-05-03	- Simplified the Analysis type.	Eduardo Julian	8	-269/+256
	- Renamed _jvm_program to _lux_program.
2016-05-03	- Removed _jvm_instanceof from the list of special forms.	Eduardo Julian	6	-90/+67

2016-05-02	- Removed _jvm_try from the list of special forms.	Eduardo Julian	7	-122/+50
	- Fixed a bug when compiling method bodies where the list of thrown exceptions did not get excluded from the safety-check when analysing the bodies.
2016-05-02	- Removed _jvm_new, _jvm_invokestatic, _jvm_invokeinterface, ↵	Eduardo Julian	6	-326/+260
	_jvm_invokevirtual and _jvm_invokespecial from the list of special forms.
2016-05-02	- Removed _jvm_getstatic, _jvm_getfield, _jvm_putstatic and _jvm_putfield ↵	Eduardo Julian	6	-153/+114
	from the list of special forms.
2016-05-01	- Removed _jvm_monitorenter, _jvm_monitorexit and _jvm_throw from the list ↵	Eduardo Julian	6	-238/+210
	of special forms.
2016-05-01	- Removed _jvm_null and _jvm_null? from the list of special forms.	Eduardo Julian	7	-99/+83

2016-05-01	- Simplified the language by eliminating coercion and macro-declaration from ↵	Eduardo Julian	6	-23/+11
	the analysis type.
2016-04-27	- Unified dozens of host operations under the _lux_host special form.	Eduardo Julian	7	-1576/+683

2016-04-26	- Now using the actual class-names for arrays of primitives, instead of the ↵	Eduardo Julian	4	-32/+49
	#Array of "something" format.
2016-04-26	- Now, the parser no longer ignores uneven-record errors.	Eduardo Julian	2	-11/+23

2016-04-26	- Made some fixes to type-inference for pattern-matching.	Eduardo Julian	4	-45/+63

2016-04-21	- Fixed a bug when lexing multi-line text.	Eduardo Julian	7	-55/+37
	- Made some minor refactorings.
2016-04-19	- Added inference for pattern-matching on tuples, records and variants.	Eduardo Julian	3	-19/+125

2016-04-18	- The compiler now takes the list of directories on which to look for source ↵	Eduardo Julian	6	-36/+36
	code during compilation.
2016-04-16	- The REPL no longer dies as soon as the first error happens.	Eduardo Julian	4	-50/+59
	- Writing "exit" in the REPL, kills it. - Fixed a bug wherein checking the presence of multiple methods with the same name being implemented, the names would overwrite each other and make some methods become impossible to check. - Fixed an error when type-checking parameterized host types. - Fixed a bug when loading the .lux files in JARs in the class-path.
2016-04-10	- Improved the loading of classes when testing whether a module is cached.	Eduardo Julian	2	-53/+56
	- Fixed a bug in which null-pointers where being compiled, rather than references to the "unit" value.
2016-03-15	- Fixed a bug in the way sub-indices for sub-tuples were being calculated ↵	Eduardo Julian	1	-2/+2
	when doing product_getLeft.
2016-03-12	- Added a REPL mode.	Eduardo Julian	8	-21/+121

2016-03-11	- Fixed the way the CLI arguments are received by _jvm_program by using the ↵	Eduardo Julian	1	-23/+9
	new format for data-structures.
2016-03-09	- Fixed a bug wherein type-variables were not getting properly cleaned and ↵	Eduardo Julian	7	-73/+120
	that was causing trouble with type-checking. - Fixed a bug in the way arguments for method-definitions were getting compiled. - Fixed a bug in the way double greater-than comparison was being compiled. - Fixed how exception signatures were getting compiled for method-definitions.
2016-03-06	- Fixed an error when parsing annotations syntax, in which the wrong ↵	Eduardo Julian	2	-2/+2
	data-structures were being returned. - Fixed an error when generating class signatures in annotations in .class files, in which the class-name was being generated instead.
2016-02-29	- VarT and ExT types are now treated as java.lang.Object.	Eduardo Julian	6	-60/+193
	- Now wrapping primitive method arguments inside objects, as Lux code always assumes everything is an object. - Now using a hand-made text lexer, that is less tolerant of escaping-cleanup errors. - When packaging the contents of dependency JARs inside the output program JAR, the compressed size of the entries is recalculated with (.setCompressedSize -1), in order to avoid errors in the JarOutputStream.