Changelog for zeolite-lang-0.17.0.0
Revision history for zeolite-lang
0.17.0.0 -- 2021-04-06
Language
-
[breaking] Adds
CharBuffer
, as a random-access, mutableChar
buffer, and makesWriteAt
a built-in type. (This is breaking becauseCharBuffer
andWriteAt
are now reserved.) -
[fix] Fixes
reduce<String,Container>
, which previously failed. -
[fix] Fixes a race condition in function dispatching that was thought to be fixed previously.
-
[behavior] Improves the memory efficiency of function calls that use 4 or fewer arguments, returns, or function params by reusing memory previously allocated for function calls.
Libraries
-
[new] Additions to
lib/util
:-
[new] Adds
Ranges
, for simple min/max determinations. -
[new] Adds
revZeroIndexed
toCounter
, to create a reversed sequence of zero-indexed container indices. -
[new] Adds
Reversed
,BySize
, andAlwaysEqual
comparators, to modifyLessThan
andEquals
behaviors for objects. -
[new] Adds
OrderH
andReadAtH
, to extendLessThan
andEquals
comparisons to simple containers.
-
-
[new] Additions to
lib/container
:-
[new] Adds
Sorting
, to support sorting of random-access containers. -
[new] Adds
TreeSet
, to manage set membership of any sortable type. -
[new] Adds
AutoBinaryTree
, to provide binary search tree (BST) functionality to custom containers that need to augment each node with additional data. -
[new] Implements
Container.size()
inSearchTree
. -
[new] Adds
KeyValueH
, to extendLessThan
andEquals
comparisons toKeyValue
.
-
-
[new] Adds
CategoricalTree
tolib/math
, as a dynamic representation of a categorical distribution over objects of any sortable type.
0.16.1.0 -- 2021-04-02
Language
-
[fix] Fixes a regression where a race condition in function dispatching would cause an error saying that a category does not implement the function that was being called.
-
[behavior] Optimizes calls to
@type
functions when made from@value
functions of the same type. -
[behavior] Optimizes
@value
initialization (e.g.,Type{ }
) when done from@type
functions of the same type.
Libraries
-
[new] Moves
Mutex
,SimpleMutex
, andMutexLock
fromlib/thread
tolib/util
. This is because they provide basic concurrency protection to objects that should not need to depend on-lpthread
.Note that this change is not breaking because
lib/util
was already a public dependency oflib/thread
. In other words, if a module already usedlib/thread
then it gotlib/util
for free. This means that modules only needing mutex functionality fromlib/thread
can instead uselib/util
. -
[new] Adds
getForward
andgetReverse
toSearchTree
, to allow the caller to iterate after looking up an entry. -
[new] Adds
checkEmpty
,checkPresent
, andcheckOptional
toTesting
inlib/testing
.
0.16.0.0 -- 2021-04-01
Language
-
[breaking] Disallows param filters in
@value interface
and@type interface
. Param filters are only used to allow procedures to call@type
and@value
functions. Previously, they were enforced in all situations where param substitution occurred, but this is no longer the case. (See further down in this list.) -
[breaking] Removes support for internal type params for values. This was an obscure feature that was never documented and was only used in some integration tests that had not been touched in over 2 years. This breaks most hand-written C++ extensions, but they can be fixed by removing the
ParamTuple
argument passed to the base-class constructor inExtValue_Foo
for C++ extensionFoo
. -
[breaking] Adds the
traverse
built-in syntax, which automatically iterates over a container that exposes the (new)Order<|#x>
@value interface
. (This is breaking becausetraverse
,Order
, andDefaultOrder
are now reserved.)traverse (container.defaultOrder() -> Foo value) { // executed once per Foo in the container }
-
[breaking] Marks
@category
member variables as read-only when initializing other member variables. Essentially, this just means that you can no longer do something like this:@category Int foo <- 1 @category Int bar <- (foo <- 2)
-
[new] Removes the requirement that type instances (e.g.,
Type<Int>
) satisfy the requirements of the type's parameter filters when not being used in a function call.concrete Type<|#x> { #x requires Foo }
In the example above,
Type<Bar>
was universally disallowed ifBar
could not be converted toFoo
. The new behavior is to enforce that restriction only when calling@type
functions, initializing values (e.g.,Type<#x>{ }
), and using the type as a param in a function call.Since
#x
is covariant in this example, you could convert aType<Bar>
to aType<any>
, even thoughany
is not a child ofFoo
. -
[new] Removes variance checks for category-level param filters. Variance was originally not checked, but then the checks were added prior to version
0.1.0.0
for obscure mathematical reasons that no longer make sense.In particular, filters such as
#x defines Equals<#x>
were not previously allowed except when#x
was invariant, but now it is allowed for all variances.The revised perspective is that param filters form a meta-type for param substitutions as inputs, with the constructed type as the output. Once the output is received, the input constraints are no longer recoverable.
Note that category-level param variance is still checked when params are used on the right side of filters for function params. This is because not doing so could result in bad argument and return types when inheriting functions.
Params also still cannot have both upper and lower bounds at the same time.
-
[new] When inferring type parameters for function calls (e.g.,
call<?>(foo)
), allowsall
to be a valid lower bound andany
to be a valid upper bound. Previously, an inferred type ofall
in covariant positions andany
in contravariant positions would cause inference to fail. -
[new] Allows
$ReadOnly[...]$
and$Hidden[...]$
for member variables in definitions ofconcrete
categories. -
[new] Makes
Argv
available to all threads. Previously, it was only available to the main thread. (Even thoughArgv
is exposed vialib/util
, this change also affects C++ extensions that useArgv
fromlogging.hpp
.) -
[fix] Fixes regression with function names used in stack traces. Previously, if a function was inherited and the types were not overridden, the category name in the trace would be that of the parent, rather than that of the one defining the procedure.
Compiler CLI
-
[breaking] Removes the default compiler mode when calling
zeolite
. The previous default was-c
, which creates a new.zeolite-module
for incremental library compilation. The default was removed because accidentally typingzeolite -f foo
instead ofzeolite -t foo
would overwrite the.zeolite-module
forfoo
rather than running its tests. -
[breaking] Updates how
$TraceCreation$
is implemented in C++. This only breaks hand-written C++ extensions that use theCAPTURE_CREATION
macro, which now must pass a string-literal category name as a macro argument, e.g.,CAPTURE_CREATION("Foo")
. -
[new] Adds the
--log-traces
option for testing mode (zeolite -t
) to capture a record of every line of Zeolite code executed when running tests. -
[fix] Fixes dependency resolution for C++ extensions that define
$ModuleOnly$
categories. -
[behavior] Improves the efficiency of call dispatching and
reduce
calls for categories that inherit a lot ofinterface
s. This is unlikely to affect performance when a category has less than 10 direct or indirect parentinterface
s.
Libraries
-
[new] Adds
ThreadCondition
tolib/thread
, which allows threads to wait for an arbitrary condition signaled by another thread. (Similar topthread_cond_t
in POSIX.) -
[new] Adds
EnumeratedBarrier
tolib/thread
, which can be use to synchronize a fixed number of threads. (Similar topthread_barrier_t
in POSIX, but with strict validation to help prevent deadlocks.) -
[new] Adds
Realtime
tolib/thread
, to allow threads to sleep based on wall time. -
[new] Adds
Counter
andRepeat
tolib/util
, to help create sequences for use with the newtraverse
built-in syntax. -
[new] Implements forward and reverse iteration to
SearchTree
inlib/container
.
Examples
- Adds
example/primes
, as a demonstration of multithreading.
0.15.0.0 -- 2021-03-24
Language
-
[breaking] Splits
Builder<#x>
built-in intoAppend<#x|>
andBuild<|#x>
to more accurately reflect its semantics. -
[breaking] Splits
ReadPosition<#x>
built-in intoReadAt<|#x>
,SubSequence
, andContainer
to be more concise, and to split up reading from subsequencing. This also includes renamingreadSize()
to justsize()
. -
[new] Adds the
Default
@type interface
, to specify a default@value
for the type. -
[fix] Fixes merging of duplicate inherited categories. Previously, if a category
Foo
was indirectly inherited multiple times then the compiler would complain about having multiple functions with the same name. -
[fix] Adds support for
^
with theBool
built-in type.
Libraries
-
[breaking] Refactors iterators in
lib/util
:-
[breaking] Makes
ReadIterator
a@value interface
. -
[breaking] Adds
AutoReadIterator
,AutoWriteIterator
, andAutoIterator
helpers to create iterators fromReadAt
andWriteAt
. (AutoReadIterator
replaces the previousReadIterator
.) -
[new] Adds
WriteAt
,WriteCurrent
,WriteIterator
, andIterator
@value interface
s.
-
-
[new] Adds
lib/container
:-
[new]
Vector
, a random-access container. -
[new]
SearchTree
, a binary search tree. -
[new]
TypeMap
, a heterogeneous key-value map of different value types. -
[new] Various
@value interface
s with container usage patterns.
-
-
[new] Adds
lib/thread
(must be built manually):-
[new]
ProcessThread
, a basic thread runner. -
[new]
SimpleMutex
andMutexLock
, for basic mutex logic. -
[new] Various
@value interface
s with thread usage patterns.
-
Compiler CLI
-
[breaking] Removes implicit dependence of C++ extensions on built-in categories
Bool
,Float
,Char
,Int
,String
, andFormatted
. Extensions that use these categories without using them in the correspondingconcrete
declaration will need to add them to therequires:
field of thecategory_source
in.zeolite-module
.category_source { source: "Extension_MyType.cpp" categories: [MyType] requires: [String,Formatted] }
-
[breaking] Cleans up
TypeInstance
caching for C++ extensions with parameters. Seelib/container/src/Extension_Vector.cpp
for an example of the new semantics, or use--templates
to create a new template.
Examples
- Removes
example/tree
. All of the corresponding code has been moved intolib/container
.
0.14.0.0 -- 2021-03-19
Language
- [breaking] Adds the
#self
meta-type. This is an implicit type param that is dynamically replaced with the type of the object a function is called on. For example, if a@value interface
function has a return type of#self
then any implementation must return an object of its own type from that function. (This is a breaking change because it reserves the param name#self
.)
Libraries
- [breaking] Removes the type parameters from
IterateForward
andIterateReverse
inlib/util
. This was made possible by the addition of the#self
meta-type.
0.13.0.0 -- 2021-03-17
Language
-
[new] Adds pragmas for local variable rules:
-
[new]
$ReadOnly[foo]$
marksfoo
as read-only in the current context following the line with the pragma. -
[new]
$Hidden[foo]$
hidesfoo
in the current context following the line with the pragma.
-
-
[fix] Fixes a latent issue with the
reduce
built-in when converting an intersection type to a union type, when one or both sides contains another nested union or intersection. This was previously fixed for compile-time checks, butreduce
was missed.
Compiler CLI
-
[breaking] Streamlines C++ extensions to cut down on boilerplate code. This is a massive change that breaks all existing C++ extensions; however, hand-written procedure definitions can be copied-and-pasted into a new extension template. Use
zeolite --templates
to regenerate new templates for C++ extensions. -
[breaking] Adds
microlens
andmicrolens-th
as dependencies, to help clean up the compiler code. This should not change the compiler's behavior.
0.12.0.0 -- 2020-12-11
Language
-
[breaking] Several more fixes for named returns:
-
[new] Allows initialization of named returns in
if
/elif
/while
predicates. Although setting a named return in a predicate has always been allowed, it previously did not count as an initialization for the purposes of checking usage in subsequent statements. -
[breaking] Fixes a bug where named returns used in
if
/elif
/while
predicates inside ofcleanup
blocks were not getting checked for initialization. This turns a potential runtime error into a compile-time error, which could break buggy code that was never going to be executed. -
[fix] Fixes a bug where a named return with a primitive type was not getting set in an explicit
return
statement. This only affected code that used primitive named returns inside of acleanup
block where the correspondingin
block used an explicitreturn
statement.@type call () -> (Int) call () (x) { cleanup { \ foo(x) // x was not getting set for this call } in return 1 // 1 was still returned from the function as expected }
This did not affect non-primitive types such as
String
and user-definedconcrete
andinterface
categories.
-
-
[breaking] Updates associativity of infix operators:
-
[breaking] Makes infix
Bool
operators&&
and||
right-associative. For example,x && y && z
used to be parsed as(x && y) && z
, whereas now it is parsed asx && (y && z)
. This better reflects the intuition of side-effects happening in order, with respect to short-circuiting. (No other operators currently short-circuit.) -
[breaking] Makes comparison operators (e.g.,
<=
) non-associative. This means that expressions such asx < y == true
will no longer compile. This is because an equivalent-looking expressiontrue == x < y
might have a different interpretation. (Note that order operators such as<
are not defined forBool
, so something likex < y < z
would have already failed previously.)
All other infix operators (including infix function calls such as
x `Math.pow` y
) remain left-associative. -
Libraries
- [new] Adds
Math.abs
forInt
absolute value tolib/math
.
0.11.0.0 -- 2020-12-09
Compiler CLI
- [breaking] Switches from
parsec
tomegaparsec
for parsing, to allow for better structuring of compile-time error messages. This should not affect syntax acceptance, but it has additional transitive dependencies that might prevent installation ofzeolite
.
Language
-
[breaking] Several
cleanup
fixes:-
[fix] Fixes a bug in
cleanup
that caused thescoped
block to be prepended to it when it was inlined in thein
block. -
[fix] Fixes a bug where
cleanup
might use an uninitialized named return when inlined at abreak
orcontinue
statement in awhile
loop. Such situations will now cause a compilation error. Previously, this could have allowed an uninitialized value to be passed around, which would only be noticed with aFunction called on null value
failure at some point. -
[fix] Fixes inlining of
cleanup
blocks whencontinue
is used inside of thein
block. Previously,cleanup
was skipped whencontinue
was used inside of the respectivein
block. -
[breaking] Disallows explicit
break
andcontinue
inside ofcleanup
blocks. Previously this was allowed (butreturn
wasn't), which led to ambiguity when thein
block contained areturn
. -
[new] Allows
cleanup
to refer to variables created at the top level of thein
block.cleanup { \ foo(bar) // bar is created below } in Type bar <- Type.create() // cleanup is inlined after this
-
0.10.0.0 -- 2020-12-05
Language
-
[breaking] Major changes to
.0rt
syntax:-
[breaking] Removes the expression argument from
success
andcrash
testcase
types. Use one or moreunittest
(see next point) instead. -
[new] Adds
unittest
keyword to allow the user to define multiple tests within a singlesuccess
testcase
.testcase "simple tests" { success } unittest checkInvariant1 { // this is a regular procedure with no return } unittest checkInvariant2 { // this is a regular procedure with no return }
This allows multiple tests to use the same setup, while still being able to track multiple test failures.
error
andcrash
testcase
types are still available. -
[breaking] Allows
testcase
to statically setArgv
(seelib/util
) using theargs
keyword. By default, only the program name is set. Previously,Argv
used whatever was passed to the test binary.testcase "with static Argv" { success args "arg1" "arg2" } unittest test { // something that requires Argv.global() }
Note that this isn't a library change; this also affects C++ extensions that use
Argv
frombase/logging.hpp
. -
[new] Adds
compiles
mode fortestcase
. This is likesuccess
except nothing is executed. (success
requires at least oneunittest
, whereascompiles
will not execute anything even ifunittest
are present.)
-
Libraries
- [new] Adds the
lib/testing
library with basic helpers for unit tests.
Compiler CLI
-
[new] Updates parsing of
.zeolite-module
to allow any ordering of the fields. Previously, the fields needed to be in a very specific order. -
[fix] Adds compile-time protection against self-referential expression macros defined in
expression_map:
in.zeolite-module
. Previously, the result was infinite recursion that could exhaust system resources. -
[breaking] Adds an error when
root
/path
in.zeolite-module
differs from the location of the respective.zeolite-module
. This is to catch both typos (e.g., copy-and-paste errors) and ambiguities.
0.9.0.0 -- 2020-11-23
Compiler CLI
-
[breaking] Major changes to C++ extensions:
-
[breaking] Corrects hole in
$ModuleOnly$
visibility in C++ extensions. Previously, C++ extensions could directly access categories defined in$ModuleOnly$
sources in dependencies. -
[breaking] Randomizes cache paths and
namespace
s for generated C++ code so that C++ extensions cannot use static#include
paths to circumvent the visibility rules of a dependency. -
[breaking] Makes
TypeInstance
s shared in generated C++ and in C++ extensions, to allow for better memory management of@type
s in the future. -
[breaking] Adds guards to C++ headers for
$TestsOnly$
categories, to prevent them from being used in non-$TestsOnly$
C++ extensions.
-
-
[fix] Fixes regression where calling
zeolite
with-p
would overwrite an existing.zeolite-module
.
Language
-
[breaking] Syntax changes:
-
[breaking] Changes the syntax for calling
@category
functions from$$
to:
, e.g.,Foo:create()
instead ofFoo$$create()
. A new error message (to be removed later) will remind the user to stop using$$
. -
[breaking] Changes the syntax for calling
@type
functions from$
to.
, e.g.,Foo.create()
instead ofFoo$create()
. A new error message (to be removed later) will remind the user to stop using$
.
-
-
[breaking] Changes to param usage:
-
[breaking] Updates param inference to be more accurate. The new algorithm will infer a param type iff the best choice is unambiguous given the expected and actual function arguments.
-
[breaking] Disallows bounding a single param both above and below, e.g., having both
#x requires Foo
and#x allows Bar
. This is primarily to prevent cycles and contradictions in the type system.Previously, the best case for such restrictions was that there are extremely few choices for
#x
, and the worst case was that it impliesBar
converts toFoo
when it isn't true, thus making#x
unusable. Having either of those requirements likely indicates a design flaw.
-
-
[breaking] Prohibits having a
$TestsOnly$
definition for a non-$TestsOnly$
category. Previously, aconcrete
category declared in a non-$TestsOnly$
.0rp
could bedefine
d in a$TestsOnly$
.0rx
, which creates a loophole that allows binaries to utilize$TestsOnly$
code. -
[fix] Fixes a latent type-checking bug that could occur when attempting to assign a value with an intersection type to a variable with a union type, if one or both types has another nested type, e.g.,
[A&B]
→[[A&B]|C]
or[[A|B]&C]
→[A|B]
. This change shouldn't cause new type-checking failures.
Libraries
- [fix] Implements
subSequence
forArgv
inlib/util
. This is required byReadPosition
but was forgotten. Calling it prior to this would cause a failure.
0.8.0.0 -- 2020-08-07
Language
-
[breaking] Makes the semantics of
cleanup
more consistent:-
[breaking] Disallows statements that modify
return
values withincleanup
blocks, i.e.,return
with values and assigning named returns. -
[new] Allows
cleanup
to access named returns that are initialized within the correspondingin
statement. Previously, access required initialization before thein
statement. -
[new] An explicit positional return (e.g.,
return 1, 2
) will assign the values to the respective named returns, if applicable. This will make the actual return values available withincleanup
.@type get () -> (Int) get () (foo) { // foo is the name of the return variable cleanup { \ bar(foo) // foo has been initialized by the time this is called } in return 1 // 1 is assigned to foo here }
-
[fix] Adds unwinding of
cleanup
blocks when used withbreak
andcontinue
, respecting loop boundaries. Previously,cleanup
was ignored.
-
-
[breaking] Skips compilation of unreachable statements. The target use-case is temporary changes to code that circumvent parts of a procedure, e.g., an early
return
for the purposes of debugging. -
[breaking] Marks statements following
break
andcontinue
as unreachable.
Compiler CLI
- [behavior] Adds a compiler warning for
public_deps
that are not required by public.0rp
sources within the same module, since they are potentially just cluttering the public namespace of the module.
0.7.1.0 -- 2020-07-13
Language
-
[new] Adds the
$SourceContext$
macro, which inserts aString
with info about source file and code location. -
[fix] Adds context to error messages related to inherited functions that have not been overridden.
0.7.0.2 -- 2020-05-21
Language
- [fix] Improves the likelihood of successful type inference when unions, intersections, or filters are combined with inheritance and type nesting.
0.7.0.1 -- 2020-05-20
Language
-
[fix] Fixes an edge-case where type-inference can be influenced by a param filter for a param with the same name in the scope that is calling the function. (For example, having
#x requires Foo
in scope while calling a function that also happens to have a param named#x
.) -
[fix] Fixes an edge-case where parameter substitution at the type level can clash with a type parameter scoped to a function. (Related to the above, but actually a separate issue discovered while solving the former.)
-
[behavior] Shows inferred-type assignments in error messages related to expressions containing inferred types.
0.7.0.0 -- 2020-05-19
Language
-
[new] Adds limited inference of type parameters in function calls. The new syntax is
call<?,Int>(foo,bar)
, where?
designates that inference is requested for the first position. -
[behavior] Reduces the memory cost of
$TraceCreation$
by avoiding storing the entire trace text.
Compiler CLI
- [behavior] Improves handling of I/O errors when calling
zeolite
, and allows compilation to continue in-r
/-R
modes if compilation of one module fails.
0.6.0.0 -- 2020-05-14
Compiler CLI
-
[behavior] Improves error messages for type mismatches and return-count mismatches.
-
[behavior] Improves the efficiency of loading metadata for dependencies.
-
[breaking] Fixes module-staleness check when running tests (
-t
). -
[breaking] Prevents binaries from being created from main categories that are defined in
$TestsOnly$
sources. (This was the original intention, but it was missed the first time around.)
Language
-
[new] Allows
@category
members inconcrete
categories to refer to each other during initialization. -
[new] Adds the
$ExprLookup[
MACRO_NAME
]$
pragma, which allows the user to define expression substitutions in.zeolite-module
. -
[new] Adds the
$NoTrace$
pragma, which skips generating stack-trace information for specific functions. This is useful for deeply-recursive functions whose full trace would not be useful. -
[new] Adds the
$TraceCreation$
pragma, which appends a trace for the creation of a value when there is a crash in one of its functions. This is useful when the crash stems from value initialization. -
[fix] Fixes parsing of the
'"'
Char
-literal.
Libraries
-
[new] Adds the
ErrorOr<#x>
category tolib/util
. This allows functions to return either the expected type or an error message. -
[new] Adds the
Void
category tolib/util
. This can be used to ignore type parameters when a value still needs to be instantiated. -
[new] Adds the
readAll
function toTextReader
(inlib/util
) to support reading an entire file at once. -
[breaking] Adds crashes when attempting to read
RawFileReader
or writeRawFileWriter
if there is a preexisting file error.
0.5.0.0 -- 2020-05-12
-
[new] Adds compiler support for pragmas, which will allow compiler features that aren't a part of the language:
-
[new]
$ModuleOnly$
pragma for.0rp
files, which limits their visibility to the modules that own them. -
[new]
$TestsOnly$
pragma for.0rp
files, which limits their visibility to.0rt
files, and.0rx
files that also use$TestsOnly$
.
-
-
[new] Adds support for global seach paths for dependencies via a plain-text file at
$(zeolite --get-path)/global-paths
. This can be used to set up a separate directory for general-use Zeolite modules. -
[breaking] Better handling of symbols from dependencies. This might break code that inavertently relied on broken visibility. In particular,
..
in a dependency path precludes it from resolving to a system module.
0.4.1.0 -- 2020-05-05
-
[new] Adds a compiler mode (
--show-deps
) to display the symbolic dependencies of a module. -
[new] Adds a compiler mode (
--fast
) to quickly compile a binary from a single source file without needing to create a module. -
[new] Adds a compiler mode (
-R
) to recursively recompile modules. -
[new] Improves thread-safety of internal code so that thread support can be safely added later. (Probably via a library.)
-
[new] Adds the
Builder
interface to makeString
concatenation more efficient. -
[new] Adds support for basic mathematical
Float
operations, such assin
,exp
, andsqrt
. -
[new] Adds support for bitwise
Int
operations. -
[fix] Fixes broken
--templates
mode when builtin types are used.
0.4.0.0 -- 2020-05-05
-
[new] Allows modules to specify custom linker flags and private include paths in
.zeolite-module
. This lets categories written in C++ depend on external libraries written in other languages. -
[behavior] Adds optimization of dependency inclusion for categories that are defined in C++ sources. This should eliminate linking in object files that are not needed by the binary.
-
[behavior] Adds checks to prevent a module from defining a category that was declared in another module.
-
[breaking] Updates the
.zeolite-module
format to require associating externally-defined categories with the source files that define them. This allows finer-grained linking of binaries and tests.
0.3.0.0 -- 2020-05-03
-
[breaking] Updates syntax for discarding all returns to use
\
instead of~
. This will allow~
to be used for bitwise-not later on. (Not to mention that~
was kind of ugly in that context.) -
[breaking] Cleans up multiple-
return
syntax by removing{}
. This applies to both thereturn
and the assignment.
0.2.0.0 -- 2020-05-03
-
[breaking] Requires that
concrete
categories defined in.cpp
files be listed asexternal
in.zeolite-module
. This allows the compiler to ensure that all categories have a definition, which helps avoid linker errors. -
[breaking] Gives
.zeolite-module
configs a proper file format. -
[new] Adds version checking to cached module data.
0.1.3.0 -- 2020-05-01
- [new] Adds support for more versions of GHC.
0.1.2.0 -- 2020-04-28
-
[fix] Fixes a parser issue with empty
{}
blocks followingscoped
. -
[behavior] Updates
cleanup
procedures to allow setting named-return values. Previously, initializing a named return incleanup
was not sufficient. -
[behavior] Updates
zeolite-setup
to unconditionally rebuild supporting libraries. Incidentally, this causes all existing user modules to be out of date.
0.1.1.0 -- 2020-04-27
-
[behavior] Set the default output path for binaries to the module's path rather than the current directory.
-
[new] Allows a base path (
-p
) with recompile mode (-r
).
0.1.0.0 -- 2020-04-27
- First version. Released on an unsuspecting world.