* math: Setting the bits of floating-point numbers
Added macros which set all 32 bits of a float, the lower 32 bits of
a double, and the upper 32 bits of a double. Some changes were made to
older code to use these macros.
* Replaced code with bitsetting macros in __tan.c3 and tan.c3
* math: tests for word macros and release notes
Tests were written for the word macros, which include getting and
setting a float with a uint and getting and setting the high or low word
of a double with a uint.
Release notes were updated to include the word setter macros.
* net/url: implement url encoding (RFC 3986)
Implement url percent-encoding and -decoding functions according to RFC
3986. Add unit tests.
Link: https://datatracker.ietf.org/doc/html/rfc3986
* net/url: ensure correct encoding of URL components
Add encoding and decoding methods to the Url struct components according
to RFC 3986.
An Url can be parsed from a String with `new_parse()` or `temp_parse()`.
The parsed fields are decoded. The only field that is not decoded is
`raw_query`. To access the decoded query values, use
`Url.query_values()`.
`Url.to_string()` will re-assemble the fields into a valid Url string
with proper percent-encoded values.
If the Url struct fields are filled in manually, use the actual
(un-encoded) values. To create a raw query string, initialize an
`UrlQueryValues` map, use `UrlQueryValues.add()` to add the query
parameters and, finally, call `UrlQueryValues.to_string()`.
---------
Co-authored-by: Christoffer Lerno <christoffer@aegik.com>
* mem: add macro to assert on memory leak in scope
Implement `mem::@assert_leak` to assert on a memory leak in the scope of
the macro body. Memory report for the leak can be disabled by setting
the boolean argument to false.
* fix: add conditional compilation flags
* Moved the code into `mem.c3` and made it a builtin.
---------
Co-authored-by: Christoffer Lerno <christoffer@aegik.com>
* Add .DS_Store to .gitignore
* Allow <= 999_999 as usec on DateTime (was < 999_999)
* Move [Tz]DateTime .format() to std::time::datetime and import only with libc
* Changed name to DateTimeFormat, prefer function over method. Move names to enum.
* Updated tests to the latest standard.
---------
Co-authored-by: Christoffer Lerno <christoffer@aegik.com>
* add std::net::url for parsing/generating URLs
* Move String.index_of_chars into std
* Fix param contract
* Idiomatic type naming, Allman formatting, slicing, document functions
* Use String.tokenize
* Don't return str_view() from freed dstring
* Change indentation to tabs
* Variable casing according to guidlelines
* Updated API and added line to the releasenotes.
---------
Co-authored-by: Christoffer Lerno <christoffer@aegik.com>
* enhance default hashing strategy for basic types
* fix
* `$defined` in a global scope should accept testing normal macros.
---------
Co-authored-by: Christoffer Lerno <christoffer@aegik.com>
If the timeout to net::poll() is -1, poll() will block until a requested
event occurs. However, in the poll() function, the Duration timeout is
converted to milliseconds (Duration is in microseconds). The conversion
involves integer arithmetics which results in a zero timeout (= -1 /
1000), i.e. poll() will return immediately and the following accept will
fail.
To fix this, only convert to milliseconds if timeout paramter is
non-negative.
Example to reproduce the error:
```
module echo;
import std::io, std::net;
fn void main()
{
TcpServerSocket server = tcp::listen("localhost", 6969, 69, REUSEADDR)!!;
server.sock.set_non_blocking(true)!!;
Poll[*] polls = {{ .socket = server.sock, .events = net::SUBSCRIBE_READ }};
if (catch net::poll(polls[..], net::POLL_FOREVER)) io::printn("poll error");
TcpSocket client = tcp::accept(&server)!!;
io::printn("OK");
}
```
Poll() will return immediately and the following tcp::accept() will fail
with an "ACCEPT_FAILED" error.
Reported-by: Alexey Kutepov <reximkut@gmail.com>
There are now nolibc definitions for the inverse cosine and inverse
sine.
More test points were added for acos, asin, and atan in the math_tests module.
This was done becuase the nolibc inverse trigonometric functions have
various branching conditions depending on the provided input value. Several
branches in these functions were neglected.
* C3_MATH feature
This feature allows the usage of noclib math files even when libc is in use.
If a nolibc symbol exists, it will be used in place of libc, otherwise
it will default to libc.
* Added MIT License notices to atan.c3
* math_tests: rewrote test_atan()
Rewrote the atan test so that analagous checks are symmetrically
performed for all possible inputs: int, float, and double. The total
number of tests has increased while reducing the total amount of
code.
* math_tests: inverse trig and inverse hyperbolic trig
Tests were written for acos, acosh, asin, asinh, and atanh.
* math: cos macro missing values::promote_int
The cosine macro can't take an integer input without this fix. You can
see that some of the other macros, like the sine macro, have this.
* math_tests: trig and exponential
Wrote tests for the trigonometric macros as well as the
exponential macro. The hyperbolic trig macros use the exponential macro
rather than any LLVM instrinsics (for now at least, LLVM 19 has the proper
compiler intrinsics).
* math: float comparison
In the math module two macros were defined to assist in comparing
floating-point numbers for a given tolerance.
The trig, hyperbolic trig, and exponential tests in the math_tests module
were updated to use these macros instead of direct comparison.
* math: fix adjoint of Matrix2
Fix the adjoint of the Matrix2x2 implementation in the math module. This
also fixes the calculation of the inverse which depends on the adjoint.
* update release notes
