Add an owned variant of WeightedChoice. #152
Conversation
Conflicts: src/libcollections/vec.rs src/libcore/fmt/mod.rs src/librustc/lint/builtin.rs src/librustc/session/config.rs src/librustc_trans/trans/base.rs src/librustc_trans/trans/context.rs src/librustc_trans/trans/type_.rs src/librustc_typeck/check/_match.rs src/librustdoc/html/format.rs src/libsyntax/std_inject.rs src/libsyntax/util/interner.rs src/test/compile-fail/mut-pattern-mismatched.rs
This gets rid of the 'experimental' level, removes the non-staged_api case (i.e. stability levels for out-of-tree crates), and lets the staged_api attributes use 'unstable' and 'deprecated' lints. This makes the transition period to the full feature staging design a bit nicer.
This gets rid of the 'experimental' level, removes the non-staged_api case (i.e. stability levels for out-of-tree crates), and lets the staged_api attributes use 'unstable' and 'deprecated' lints. This makes the transition period to the full feature staging design a bit nicer.
This adds the int_uint feature to *every* library, whether or not it needs it.
Conflicts: src/test/compile-fail/borrowck-move-out-of-overloaded-auto-deref.rs src/test/compile-fail/issue-2590.rs src/test/compile-fail/lint-stability.rs src/test/compile-fail/slice-mut-2.rs src/test/compile-fail/std-uncopyable-atomics.rs
Fixes #16072
Fixes #16072 r? @huonw
Lets them build with the -dev, -nightly, or snapshot compiler
* `core` - for the core crate * `hash` - hashing * `io` - io * `path` - path * `alloc` - alloc crate * `rand` - rand crate * `collections` - collections crate * `std_misc` - other parts of std * `test` - test crate * `rustc_private` - everything else
* `core` - for the core crate * `hash` - hashing * `io` - io * `path` - path * `alloc` - alloc crate * `rand` - rand crate * `collections` - collections crate * `std_misc` - other parts of std * `test` - test crate * `rustc_private` - everything else
Conflicts: mk/tests.mk src/liballoc/arc.rs src/liballoc/boxed.rs src/liballoc/rc.rs src/libcollections/bit.rs src/libcollections/btree/map.rs src/libcollections/btree/set.rs src/libcollections/dlist.rs src/libcollections/ring_buf.rs src/libcollections/slice.rs src/libcollections/str.rs src/libcollections/string.rs src/libcollections/vec.rs src/libcollections/vec_map.rs src/libcore/any.rs src/libcore/array.rs src/libcore/borrow.rs src/libcore/error.rs src/libcore/fmt/mod.rs src/libcore/iter.rs src/libcore/marker.rs src/libcore/ops.rs src/libcore/result.rs src/libcore/slice.rs src/libcore/str/mod.rs src/libregex/lib.rs src/libregex/re.rs src/librustc/lint/builtin.rs src/libstd/collections/hash/map.rs src/libstd/collections/hash/set.rs src/libstd/sync/mpsc/mod.rs src/libstd/sync/mutex.rs src/libstd/sync/poison.rs src/libstd/sync/rwlock.rs src/libsyntax/feature_gate.rs src/libsyntax/test.rs
Conflicts: mk/tests.mk src/liballoc/arc.rs src/liballoc/boxed.rs src/liballoc/rc.rs src/libcollections/bit.rs src/libcollections/btree/map.rs src/libcollections/btree/set.rs src/libcollections/dlist.rs src/libcollections/ring_buf.rs src/libcollections/slice.rs src/libcollections/str.rs src/libcollections/string.rs src/libcollections/vec.rs src/libcollections/vec_map.rs src/libcore/any.rs src/libcore/array.rs src/libcore/borrow.rs src/libcore/error.rs src/libcore/fmt/mod.rs src/libcore/iter.rs src/libcore/marker.rs src/libcore/ops.rs src/libcore/result.rs src/libcore/slice.rs src/libcore/str/mod.rs src/libregex/lib.rs src/libregex/re.rs src/librustc/lint/builtin.rs src/libstd/collections/hash/map.rs src/libstd/collections/hash/set.rs src/libstd/sync/mpsc/mod.rs src/libstd/sync/mutex.rs src/libstd/sync/poison.rs src/libstd/sync/rwlock.rs src/libsyntax/feature_gate.rs src/libsyntax/test.rs
Add categories to Cargo.toml
i128 support
Derive Debug for distibutions
Add a note about `OsRng` blocking in early init.
Add note that shuffle uses Fisher–Yates shuffle which is unbiased
For example, `sample(r, b"ATGC", 4)` always returns `[b'A', b'T', b'G', b'C']`.
Clarify reservoir sampling docs
|
There is an alternative fix in #151. |
The change to WeightedChoice is debatable, but since the argument must normally be constructed specially anyway, the reference model makes little sense. See issues rust-random#90, rust-random#142 and PR rust-random#152.
|
There are a couple of other alternatives: make the current I've written code to do similar jobs a few times before, using floating-point numbers and iterative search or Recommendation: store the items and weights in separate arrays — or even leave storage of the items to users and only store weights. This means the See also this issue: dhardy#82 |
|
Sorry this PR got so suddenly closed. The cause was that we cleaned up the git history (#350) and forced-pushed it to master. Now there was a difference of about 30.000, and I suppose GitHub just gave up... This PR, together with two others, stayed open for so long with little attention because it touches on an area of Rand that needs some design first. That is now slowly getting explored in dhardy#82. |
|
Can we have a slightly higher priority on this? |
@usamec I can believe it took quite some Googling to find this PR 😄. As you maybe know development in this repro only started up again a couple of months ago, ane we have been hard at work getting the foundation right. I hoop very soon we will see the 0.5 release with the changes.
Two comments about weighted sampling: dhardy#82 (comment) and dhardy#82 (comment). Do you need weighted sampling with replacement, or without? For weighted sampling with replacement I still need to organise my notes and make a post. The current implementation of For now I'm afraid it is best write/copy an implementation from scratch (or help us out?). |
|
@pitdicker Thanks for reply. |
Fixes #90, fixes #142.