Required a few changes to APIs, namely janet_root_fiber()
to get topmost fiber that is active in the current scheduler.
This is distinct from janet_current_fiber(), which gets the bottom
most fiber in the fiber stack - it might have a parent, and so cannot
be reliably resumed.
This is the kind of situation that makes symmetric coroutines more
attractive.
This way we can support fewer build configurations. Also, remove
all undefined behavior due to use of memcpy with NULL pointers. GCC
was exploiting this to remove NULL checks in some builds.
This means we can add new properties to abstract types without
breaking old code. We can also make simple abstract types without
needing to add many NULL fields to the type.
This changes the implementation of the `next` function which
is now used to implement each. This let's us iterate over
more types, not just tables and structs.
In some cases, one might want to disable what is currently
SipHash for speed / better security mechansims. For example, using
red black trees for caches rather than hash tables.
The hash key still needs to randomly initialized
for the security advantage, but this patch is a
step closer to avoiding hash based DOS.
Further work may including exposing the raw hash
function for use by abstract types who also choose to
implement hash.
If JANET_ENTRY_NAME is defined, we are compiling into a single binary.
In this case, we don't want to define the config symbol multiple times
with same name, as this causes the linker error.
While C functions are not re-entrant, signaling from a C function
can be used to implement async returns. When resuming a fiber that
signalled from within a C function, the fiber is started after the
instruction that emitted the signal. The resume argument is used
as the return result from the c function.
This unifies equality and comparison checking. Before, we had
separate functions and vm opcodes for comparing general values vs.
for comparing numbers, where the numberic functions were polymorphic and
had special cases for handling NaNs. By unfiying them, abstract types
can now better integrate with other number types and behave as keys.
For now, the old functions are aliased but will eventually be removed.
Gives more control over unmarshalling
abstract types. This should also
make it possible/easy to write abstract types that cannot
cause unmarshal to segfault.
This makes the names of the opcodes match their implied functionality.
We also rename the C functions to match the opcodes and source level
functionality.
Rather than messing with janet_core_dictionary, we
instead cache the core enevironment, and pull out the
needed tables from there. This is more flexible, more correct, and
also exposes janet_resolve_core, which can be easily used from the C
API.
The janet_get_permissive function implements the core semantics
of the 'get' function. The original janet_get implements the semantics of
the 'in' function and also the OP_GET opcode. This slight oddity is
to avoid a backwards incompatible change.
A finalizer can be attached to scratch allocations efficiently at any point in
it's lifecycle via janet_sfinalizer. Care was taken to keep allocations aligned
with platform alignment requirements.
A big drawbacks to this approach is the waste of up to 16 bytes per scratch
allocation in the case the scratch memory does not require a finalizer.
The new RNG wraps up state for random number generation, so
one can have many rngs and even marshal and unmarshal them.
Adds math/rng, math/rng-uniform, and math/rng-int.
Also introduce `in` and change semantics for
indexing out of range. This commit enforces stricter
invariants on keys when indexing via a function call
on the data structure, or the new `in` function.
The `get` function is now more lax about keys, and will
not throw an error when a bad key is used for a data structure, instead
returning the default value.
The print family of functions now writes output
to an optional buffer instead of a file bound to :out.
This means output can be more easily captured an redirected.
This should be friendlier to most users. It does, however, mean
we lose range information. However, range information could be
recovered by re-parsing, as janet's grammar is simple enough to do this.
Also adds the janet_lengthv API call. This is
needed because janet_length returns a 32 bit integer, where
as lengthv lets us return larger values (useful for typed arrays).
janet_mcall is an api function that should make it easier to call
a janet method from C code. It shares a similar signature with
janet_call.
A keyword reference only counts as visited if we have
it as cached in the memoized->table, and we know it was
originally referenced from the same grammar table. If these
two conditions are true, then compilation must work correctly.
Also add janet_table_get_ex.
Mostly changes to cook and jpm. Also some
code for file associations in the windows installer, and
adding the :linux value from os/which (instead of just :posix).
This allows better stacktraces when manually intercepting
signals to clean up resources. Also allows functionality
from Common Lisp's unwind-protect, such as calling cleanup code
while unwindinding the stack, restarting on certain signals, and
just in general having more control over signal and signal propagation.
Also fix a bug encountered while implementing with-resource in the
compiler. Desturcturing arguments that were not the last argument
would often result in bad code generation, as slots used to destructure
the earlier arguments would invalidate the later parameters. This is
fixed by allocating all named parameters before doing any destructuring.
This will allow some one constructing an abstract to
only make it visible to the garbage collector after it
is in a valid state. If code in the constructing cfunction
panics before janet_abstract_end is called, the GC will not try
to mark the incomplete abstract type. This is often not needed through
careful programming, but should work well.
This lets us build a smaller binary. The minimal tested
binary on x86-64 (with -Os, -s, and all options that shrink binary size
turned on) is about 240 kB.
These functions made it very easy to create memory
leaks, and are better replaced with functions in vector.h or
simply using non-stack allocated arrays.
This makes it easier to document functions that
take keyword arguments and also prevents some allocations
with these functions. Before, this was possible via normal
variadic functions but created an intermediate tuple, and
the generated docstrings did not document the keys.
This should help address #81. Also hide janet_exit
and janet_assert, as they are really meant for internal usage.
I have not verified that this yet actually works with Rust's
bindgen.
64 bit nanboxing is kind of sketchy on non x86 architectures.
32 bit architectures seem to work better as the 32 implementation
doesn't rely on the format of the address space and layout of
double's in memory.
- Allow passing a table to fibers, which make fiber level scope easier.
- Add fiber/getenv, fiber/setenv, dyn, and setdyn
- Remove meta, *env*, and *doc-width*
- Some functions changed dignatures, and no longer take an env
Rather than edit the Makefile or the janet.h header yourself, use
janetconf.h to configure builds. This has the benefit of making it
easier to configure janet in a persitent but easy way.
A lot of refactoring larger integer types. Fix a number
of casting errors, but mostly rename things. Also try to
limit use of template-like macros as they bloat the binary
if not used in moderation. We were able to reduce the size of
typed array code as well by using a single view types.
We moved the literals true and false into one tag
type, so we an extra tag for raw pointer types
(light userdata). These can be used from the C API via
janet_wrap_pointer and janet_unwrap_pointer.
Allow overriding functions in the core libray to provide better
functionality on startup. Used to include our getline function in
the repl but use a simpler version in the core library.
Instead of a int32_t as the length argument, use
size_t to match up better with typearray.c and probably
most idiomatic C libraries.
Janet uses int32_t for length internally for consistency, space
efficiency, ability to fit int32_t in double, and various
other reasons.
Typed arrays used size_t in serialization: C APIs will
also often use it, so it makes sense to add first class support
for it rather than assume it will will fint into an integer.
These changes should quiet some visual studio warnings.
Also make some spacing more consistent.
Leave in issues with calling memcpy with size=0. If these
become a problem, will probably add a janet_memcpy as memcpy
is used so much in the code without 0 checks.
If a comment is not followed by a newline character, then
we got a false parse error. This is because the comment
state is left on the parse stack when we finished parsing, and
since the parse stack was not emtpy, we assumed an error.
This commit adds the parser/eof function, which lets the parser know
that an eof was reached. Before, we simply added a fake newline
character in some cases, and in the case of reading a file, we did
nothing, hence the bug.
Before, we used a local setjmp/longjmp for error handling.
Using janet_panic means errors can be more easily expressive and
code can be smaller.
However, we still need to make vector memory get gc collected, as
panics can cause the runtime to skip janet_v_frees.
A consistent style should help with contributors and
readability. We use astyle as the formatter as can make a pretty
good approximation of the current style and my preferred style.
Astyle can be found at http://astyle.sourceforge.net/astyle.html
This should speed up start time and reduce malloc/free
usage to about 15% of what is what previously for startup.
The current cost is slightly larger binary as the representaion
of the image is currently less compact than source code.
There was an implementation for stacktraces in both
run.c and in core.janet, status-pp. The commit removes
the one in core.janet in favor of the C based stacktrace, which
is exposed via debug/stacktrace. Lots of reshuffling of run-context
ensued as well, which resulted in an api that is a bit cleaner.
to make janet function calls easier and faster from C (still
needs an object pool for fibers, though). Fix bug in scan-number
and add many more peg tests.
Error handling is implemented with setjmp/longjmp so code
can be more concise. This required a very large but straight forward refactor for all
of the libraries.
remove some complexity and unexpected behavior around numbers in
general as all numbers are the same number type, IEEE 754 double
precision numbers. Also update examples and tests, some of which were
out of date.
Some more testing may be needed for new changes to numbers.
