89aac82568
Fix dead link
2020-08-28 21:37:24 +02:00
5bf110b980
Don't use deprecated import path
2020-08-28 21:37:24 +02:00
a393403d08
Update outdated links
2020-08-28 21:37:24 +02:00
42247d8616
Restrict rand::rngs::adapter to std
...
See #911 .
2020-08-28 18:47:54 +02:00
fbcfa2261d
Fix chacha performance
...
This reverts a change from #979 that introduced a massive performance
regression by disabling SIMD for `rand_chacha`.
Likely, the behavior of `ppv-lite86` should rather be fixed. This change
serves as a workaround until a fixed version is available.
Fixes #1017 .
2020-08-28 18:07:38 +02:00
8d0e1f4d48
StdRng: Switch to ChaCha12 (#1028 )
2020-08-28 16:29:51 +01:00
2c6c6b1be1
Update changelog
2020-08-28 15:31:56 +02:00
9577e0b1c9
Slightly improve test
2020-08-28 15:31:56 +02:00
5bd9dcf642
Delete dead code
2020-08-28 15:31:56 +02:00
e88f2cba06
Restore unsafe fill_via_chunks implementation for performance
2020-08-28 15:31:56 +02:00
9ccdffd095
Simplify code
2020-08-28 15:31:56 +02:00
dae0d96286
Check source size and prefer chunks_exact
2020-08-28 15:31:56 +02:00
ad08968b00
Use chunks_exact_mut for slightly better performance
...
The results from master, using unsafe code:
```
gen_bytes_chacha12: 2,733,838 ns/iter (+/- 181,694) = 374 MB/s
gen_bytes_chacha20: 4,339,602 ns/iter (+/- 237,793) = 235 MB/s
gen_bytes_chacha8: 1,918,279 ns/iter (+/- 103,581) = 533 MB/s
```
The results of the new code using `chunks_exact_mut` (this commit):
```
gen_bytes_chacha12: 3,049,147 ns/iter (+/- 220,631) = 335 MB/s
gen_bytes_chacha20: 4,645,772 ns/iter (+/- 269,261) = 220 MB/s
gen_bytes_chacha8: 2,214,954 ns/iter (+/- 1,745,600) = 462 MB/s
```
The results of using `chunks_mut` (before this commit):
```
gen_bytes_chacha12: 3,492,109 ns/iter (+/- 164,638) = 293 MB/s
gen_bytes_chacha20: 5,087,706 ns/iter (+/- 249,219) = 201 MB/s
gen_bytes_chacha8: 2,700,197 ns/iter (+/- 524,148) = 379 MB/s
```
2020-08-28 15:31:56 +02:00
4bb5cfa485
Simplify macro
2020-08-28 15:31:56 +02:00
f87ef95344
Fix fill_via_chunks and add tests
2020-08-28 15:31:56 +02:00
089a63878e
WeightedIndex: Fix serde test
2020-08-28 15:31:56 +02:00
786b95fa10
Simplify tests to address review feedback
2020-08-28 15:31:56 +02:00
1589af3543
Try to fix endianess issue
2020-08-28 15:31:56 +02:00
812e7b35f8
Simplify cfg logic
...
This is possible thanks to `alloc` being implied by `std` builds since
Rust 1.36.
2020-08-28 15:31:56 +02:00
80801fb78e
Avoid indexing
2020-08-28 15:31:56 +02:00
c3f0a30ddd
Drop some unsafe code from rand_core
...
The necessary standard library functions were stabilized with Rust 1.34.
Our MSRV is 1.36.
2020-08-28 15:31:56 +02:00
dd4b018b23
Reenable alloc tests for MSRV
2020-08-28 15:31:56 +02:00
7555747c16
Bump MSRV to 1.36
2020-08-28 15:31:56 +02:00
9a3f2f4e7a
sample_weighted: Make sure the correct IndexVec is generated
...
Also add some tests.
2020-08-27 18:21:07 +02:00
29e78c7b75
Get rid of "partition_add_index" feature
...
It offers nothing over the "nightly" feature and makes testing more
complicated.
2020-08-27 18:21:07 +02:00
437f789b06
sample_weighted: Use less memory for length <= u32::MAX
2020-08-27 18:21:07 +02:00
519eb70d43
Address review feedback
2020-08-27 18:21:07 +02:00
45a2c390ca
sample_weighted: Return early if amount is 0
2020-08-27 18:21:07 +02:00
158df6c34a
Undo reformatting
2020-08-27 18:21:07 +02:00
57d0cb0ad4
Don't use IntoIter when unnecessary and fix complexity documentation
2020-08-27 18:21:07 +02:00
330909be42
Don't use rustc features newer than 1.32
2020-08-27 18:21:07 +02:00
a310f9720c
Fix feature flags and use core for nostd compatibility
2020-08-27 18:21:07 +02:00
8890183f13
Mark new tests with the correct feature flags
2020-08-27 18:21:07 +02:00
9b4df3bb0b
Add choose_multiple_weighted, tests, and benchmarks
2020-08-27 18:21:07 +02:00
cb493927e6
Use doc(cfg) to annotate feature-gated items ( #1019 )
...
* Use doc(cfg) to annotate feature-gated items
* Fix block RNG example
* Formatting (partial rustfmt run)
2020-08-27 10:20:28 +01:00
27757fb229
Update changelog for rand_distr 0.3.0 ( #1024 )
...
Co-authored-by: Diggory Hardy <git@dhardy.name >
2020-08-27 08:39:09 +01:00
0f933f9c71
Correctly spell "implementation"
2020-08-06 18:29:01 +02:00
bbb0dff53c
Fix benchmarks
2020-08-05 13:25:25 +02:00
0610312c89
Alphanumeric samples bytes instead of chars
...
The corresponds more closely to the internally used types and can be
easily converted to a `char` via `From` and `Into`, while being more
flexible to use.
This is a breaking change.
2020-08-05 13:25:25 +02:00
7f321fa21f
Impl Distribution<u8> for Alphanumeric
...
Sampling a random alphanumeric string by collecting chars (that are known to be ASCII) into a String involves re-allocation as String is encoding to UTF-8, via the example:
```rust
let chars: String = iter::repeat(())
.map(|()| rng.sample(Alphanumeric))
.take(7)
.collect();
```
I wanted to get rid of the clearly unnecessary re-allocations in my applications, so I needed to be able to access to the ASCII characters as simple bytes. It seems like that was already what was going on inside Alphanumeric however, it was just internal.
This PR changes the `Distribution<char>` impl to provide `u8`s (which it generates internally) instead, and implements the previous `Distribution<char>` using it. One could then, for example, do this:
```rust
let mut rng = thread_rng();
let bytes = (0..7).map(|_| rng.sample(ByteAlphanumeric)).collect();
let chars = unsafe { String::from_utf8_unchecked(bytes) };
```
2020-08-05 13:25:25 +02:00
dca9cb53eb
Avoid to give advice on password generation
2020-08-04 17:29:46 +02:00
e0395fd713
Condense advice on password generation
2020-08-04 17:29:46 +02:00
2da40a3691
Remarks on passwrod generation
...
I suppose the code should go elsewhere.
2020-08-04 17:29:46 +02:00
321ae3962b
fix build without std/alloc features
2020-08-02 02:39:46 +02:00
a75317ea4a
fix test
2020-08-02 02:39:46 +02:00
7781769f04
remove weighted module, re-export WeightedAliasIndex to top level
2020-08-02 02:39:46 +02:00
5cebc0c083
Rename alias_method::WeightedIndex to WeightedAliasIndex
2020-08-02 02:39:46 +02:00
c4dda363e3
Update changelogs ( #1010 )
2020-08-01 18:43:16 +01:00
98b915feed
Change gen_range to accept Range types
...
This adds support for `Rng::range(a..b)` and `Rng::range(a..=b)`,
replacing `Rng::range(a, b)`. `a` and `b` must now implement
`PartialEq`.
This is a breaking change. In most cases, replacing
`Rng::gen_range(a, b)` with `Rng::gen_range(a..b)` should be enough,
however type annotations may be necessary, and `a` and `b` can no longer
be references or SIMD types.
SIMD types are still supported by `UniformSampler::sample_single`.
Some clippy warnings were fixed as well.
2020-08-01 16:09:50 +02:00
17651c0855
Add UniformSampler::sample_single_inclusive
...
This includes a specialized implementation for integers.
2020-08-01 16:09:50 +02:00
Vinzent Steinberg
Zachary Neely
Diggory Hardy
Tamir Duberstein
qoh
Jeff Burdges
Артём Павлов [Artyom Pavlov]
Benjamin Saunders