* first pass at signal support for node backend
* change signal values to int's
* implements defaultSignal
* return -1 as expected by calling API if any error is raised by nodejs runtime
* finishes signal support for nodejs
* extract repetitive foreign import identifiers
* fix comments
As discussed with edwin, let's get rid of the external TT type.
There is no way to get your hands on a TT value anyway so this
should not change anything.
Put the `RWST` argument to be the last one. This makes such functions
to be easier used in point-free compositions and to be easily
interchangeable with existing `runStateT`-like functions.
`Given` with `Always` from Idris 1 library are completely overridden by
`IsYes` and `ItIsYes` respectively, which have a more common naming.
This, however, may break some very old code (fixed by a trivial rename).
* RefC backend improvements
1. OnCollect had the wrong number of arguments. The code creator expects
3 arguments, but onCollect in prim.h expected 4 arguments. The first of which
was an erased arguments. That is now fixed.
2. OnCollect did not call `newReference` when creating a new reference to the pointer
and the freeing function
3. OnCollect and OnCollectAny still had a spurious printf statement
Those issues have been fixed, the test case can be found in
tests/refc/garbageCollect
4. The IORef mechanism expects that the %World token will be passed around
consistently. This is not the case. States in Control.App make use of
IORefs, but the function created from Control.App.prim_app_bind
had the world token erased to NULL.
Now, IORefs are managed using a global variable,
IORef_Storage * global_IORef_Storage;
referenced in cBackend.h, defined in the created .c file, and set to NULL
in main();
5. While multithreading and forking is still not supported, compiling a program
that makes use of Control.App demands a C implementation of prim_fork.
Files support/refc/threads.c and support/refc/threads.h provide a
dummy implementation for it, so that Control.App programs compile and run.
A test for these 2 issues is given in tests/refc/issue2424
* format changes
to make the linter happy
* format changes
to make the linter happy
* format changes
to make the linter happy
* spelling mistake braket -> bracket
Co-authored-by: Volkmar Frinken <volkmar@onutechnology.com>
* Allow functions to be marked for foreign export
This relies on the backend knowing what to do with such things, but the
general idea is to mark them with '%export "backend:exportedname"' then
'getCompileDataWith', given a back end, will search for every function
that needs to be exported, as well as every function starting from the
expression to be compiled. This will allow Idris functions to be called
from other languages, where a backend supports it.
This is hard to set up a test case for, for the moment, since no
backends actually do anything with it. So consider it a bit of a
placeholder for now.
* Add missing clause to Eq FnOpt
Thanks to @buzden
The scoping of the utility function meant that Idris treats `foldl f z (xs :< x)` and `f (foldl f z xs) x` as two different terms, making proving things about it difficult.
* support for system command via node backend.
* Add env var set/unset
* fix env unset function
* Update libs/base/System.idr
* modify system test to cover node and chez.
* Add base tests for env get/set
* [ fix ] Fix returned status of the `system` function
* [ re #238 ] Fix program error condition of `git diff` call in `Golden`
According to documentation, not only negative exit code means error
* contrib library unused import removal
* remove a few unused imports.
* another round of unused import removal
* another round of unused import deletion.
* another round of unused import deletion.
This is (for once) not a breaking changes, instead backends will need to opt in to this change, using the utilities in Compiler.NoMangle. See the js backend for an example of how to do this.
This is the first step to being able to use idris to create libraries usable by other languages.
* Only normalise a search goal if it's fast
While we do end up normalising it anyway on success, there might be
things blocking it that make the intermediate terms very big, so only do
it speculatively to see if it's quick.
* Get information about names in reflection
Currently this is only whether it's a function, or data or type
constructor. I expect more may be useful/possible.
The prelude interfaces that have default definitions for all of
their fields are declared total so that users are forced to think
about meeting the minimal requirements for an implementation to be
valid.
* deprecate Data.Nat.Order.decideLTE
* Add properties for LTE/GTE that produce the difference.
* remove deprecated function now that it is available in the base library.
* remove two deprecated lines.
* remove module deprecated since v0.4.0
* fix prelude reference to renamed primitive.
* finish removing Data.Num.Implementations
* remove deprecated dirEntry function.
* remove deprecated fastAppend. Update CHANGELOG.
* replace fastAppend in test case
* replace fastAppend uses in compiler.
* remove new properties that weren't actually very new.
* Implemented %noinline
* Removed trailing spaces.
* Added missing case in Reify FnOpt
* Added error message when both %inline and %noinline are set.
* Added test.
* Changed from perror to error
* Case tree/coverage checking shortcuts
We were calculating some things we didn't need - we can stop computing
the type of a case operator when we know the head, because that's all we
need for coverage checking. We can also abandon checking a left hand
side for coverage purposes if we encounter an empty type. Both of these
can save quite a bit of time in complex cases.
* Normalisation heuristic for pattern variables
If they get bit, fully normalise (like we do with case types) since it's
likely a big term with lots of applications will normalise a lot.
* [ refactor ] user proper int types for Constant
* [ cleanup ] declare standalone TTC implementations for BitsN/IntN
Rather than doing the casting inline, have the (en/de)coding all
side by side in one place
* [ cleanup ] remove duplicated code
Co-authored-by: Guillaume ALLAIS <guillaume.allais@ens-lyon.org>
* Update version numbers and bootstrap scheme
* Use wall clock time for search timeouts
That was always the intention in any case, rather than the process time.
Instead of having UN & RF (& Hole in the near future & maybe even
more later e.g. operator names) we have a single UN constructor
that takes a UserName instead of a String.
UserName is (for now)
```idris
data UserName : Type where
Basic : String -> UserName -- default name constructor e.g. map
Field : String -> UserName -- field accessor e.g. .fst
Underscore : UserName -- no name e.g. _
```
This is extracted from the draft PR #1852 which is too big to easily
debug. Once this is working, I can go back to it.
* Add function that checks whether a file is a terminal device.
* support isTTY function for NodeJS backend.
* don't accidentally interpret 'false' string as truthy number
* less code duplication.
integerToNat is only equal to `believe_me` at runtime, not at compile
time. You'd believe it cannot be a problem given that the implementation
of `Cast` is not exported but the REPL reduces everything.
* [ breaking ] remove parsing of dangling binders
It used to be the case that
```
ID : Type -> Type
ID a = a
test : ID (a : Type) -> a -> a
test = \ a, x => x
```
and
```
head : List $ a -> Maybe a
head [] = Nothing
head (x :: _) = Just x
```
were accepted but these are now rejected because:
* `ID (a : Type) -> a -> a` is parsed as `(ID (a : Type)) -> a -> a`
* `List $ a -> Maybe a` is parsed as `List (a -> Maybe a)`
Similarly if you want to use a lambda / rewrite / let expression as
part of the last argument of an application, the use of `$` or parens
is now mandatory.
This should hopefully allow us to make progress on #1703
* Stub for future 'identity' optimisation
I plan to add this later, but I'm using for now for
NaturalToInteger and IntegerToNatural
* Refactor `%builtin`
fixes#1799
- automatically optimise all Natural shaped things
- NaturalToInteger and IntegerToNatural now use
new `Identity` flag (internal use only for now)
which signals the function is identity at runtime
* Use NaturalToInteger and IntegerToNatural for Nat and Fin
Also define show fin in terms of finToInteger, for speed
* Fix name handling for %builtin
* [ tests ] fixes + #1799
* remove %builtin from libs
Add back after next version
* Use resolved names where convenient
The previous definition means it won't reduce until it's applied to
4 arguments which may have detrimental effects: ``f `on` fst`` would
for instance stay blocked, with some implicit arguments of the form
`DPair a b`. This means that `b` appears in a negative position in
the expression which may lead to positivity checking rejecting a
datatype defined using `on`!
I have decided to leave `g` and `f` on the LHS because I expect `on`
to be used either:
1. partially applied to two arguments
2. in a section if only applied to 1 and sections get eta-expanded
by the parser so that's fine.
Turns out that `Smaller` and `LT` won't unify because
1. the instance Sized Nat is not publicly exported
2. Smaller, and LT are stuck until fully applied
The given changes make that go away.
I can't make sense of this code, it seems to try to convert the
case function corresponding to `let (a, b) = f n in ...` into a
different case function where `f n` and `(a, b)` have been unified.
But if `f n` is a bona fide stuck computation, there's no chance of
this happening.
Just turning this off solves the #1782 and only breaks one function
in the whole of the idris2 repo (I would have expected our current
termination oracle to be too weak to detect it as valid anyway!)
and one in frex (which, again, should not have been seen as terminating).
Also fixes#1460
In the `MkFix : f (Fix f) -> Fix f` example, using `Erased` for `f`
makes the type reduce to `[__] (Fix [__]) -> Fix [__]` and because
`[__] e1 ... en` computes to `[__]`, we end up with `[__] -> Fix [__]`
which does not reference `Fix` anymore.
A few proofs have been rewritten, a few unnecessary cases cut, and
lots of unnecessary "explicit implicits" have been cut. Probably these
implicits were required when the code was initially written, and
inference has improved since then.
`testInDir dir ...` lists all directories in `dir` which contains
`run` files, and such directories are considered tests.
This is done to make test addition/maintenance cheaper.
Convert some test directories to `testInDir`, but not all of them
because
* some directories are listed in several test groups
* other directories are have some tests disabled
These entries returned by `readdir` are legacy of Unix API, we don't
really need them. Most APIs do not return them (for example Java
`Files.newDirectoryStream` or Python `os.listdir`).
* add `nextDirEntry` which returns `Maybe String`, so `Nothing` on
the end of directory unlike `dirEntry` which returns unspecified error
on the end of directory
* `dirEntry` is deprecated now, but not removed because compiler depends on it
* native implementation of `dirEntry` is patched to explicitly reset `errno`
before the `readdir` call: without it end of directory and error were
indistinguishable
* test added
