# Description
This PR introduces a `ByteStream` type which is a `Read`-able stream of
bytes. Internally, it has an enum over three different byte stream
sources:
```rust
pub enum ByteStreamSource {
Read(Box<dyn Read + Send + 'static>),
File(File),
Child(ChildProcess),
}
```
This is in comparison to the current `RawStream` type, which is an
`Iterator<Item = Vec<u8>>` and has to allocate for each read chunk.
Currently, `PipelineData::ExternalStream` serves a weird dual role where
it is either external command output or a wrapper around `RawStream`.
`ByteStream` makes this distinction more clear (via `ByteStreamSource`)
and replaces `PipelineData::ExternalStream` in this PR:
```rust
pub enum PipelineData {
Empty,
Value(Value, Option<PipelineMetadata>),
ListStream(ListStream, Option<PipelineMetadata>),
ByteStream(ByteStream, Option<PipelineMetadata>),
}
```
The PR is relatively large, but a decent amount of it is just repetitive
changes.
This PR fixes#7017, fixes#10763, and fixes#12369.
This PR also improves performance when piping external commands. Nushell
should, in most cases, have competitive pipeline throughput compared to,
e.g., bash.
| Command | Before (MB/s) | After (MB/s) | Bash (MB/s) |
| -------------------------------------------------- | -------------:|
------------:| -----------:|
| `throughput \| rg 'x'` | 3059 | 3744 | 3739 |
| `throughput \| nu --testbin relay o> /dev/null` | 3508 | 8087 | 8136 |
# User-Facing Changes
- This is a breaking change for the plugin communication protocol,
because the `ExternalStreamInfo` was replaced with `ByteStreamInfo`.
Plugins now only have to deal with a single input stream, as opposed to
the previous three streams: stdout, stderr, and exit code.
- The output of `describe` has been changed for external/byte streams.
- Temporary breaking change: `bytes starts-with` no longer works with
byte streams. This is to keep the PR smaller, and `bytes ends-with`
already does not work on byte streams.
- If a process core dumped, then instead of having a `Value::Error` in
the `exit_code` column of the output returned from `complete`, it now is
a `Value::Int` with the negation of the signal number.
# After Submitting
- Update docs and book as necessary
- Release notes (e.g., plugin protocol changes)
- Adapt/convert commands to work with byte streams (high priority is
`str length`, `bytes starts-with`, and maybe `bytes ends-with`).
- Refactor the `tee` code, Devyn has already done some work on this.
---------
Co-authored-by: Devyn Cairns <devyn.cairns@gmail.com>
# Description
On 64-bit platforms the current size of `Value` is 56 bytes. The
limiting variants were `Closure` and `Range`. Boxing the two reduces the
size of Value to 48 bytes. This is the minimal size possible with our
current 16-byte `Span` and any 24-byte `Vec` container which we use in
several variants. (Note the extra full 8-bytes necessary for the
discriminant or other smaller values due to the 8-byte alignment of
`usize`)
This is leads to a size reduction of ~15% for `Value` and should overall
be beneficial as both `Range` and `Closure` are rarely used compared to
the primitive types or even our general container types.
# User-Facing Changes
Less memory used, potential runtime benefits.
(Too late in the evening to run the benchmarks myself right now)
# Description
Does some misc changes to `ListStream`:
- Moves it into its own module/file separate from `RawStream`.
- `ListStream`s now have an associated `Span`.
- This required changes to `ListStreamInfo` in `nu-plugin`. Note sure if
this is a breaking change for the plugin protocol.
- Hides the internals of `ListStream` but also adds a few more methods.
- This includes two functions to more easily alter a stream (these take
a `ListStream` and return a `ListStream` instead of having to go through
the whole `into_pipeline_data(..)` route).
- `map`: takes a `FnMut(Value) -> Value`
- `modify`: takes a function to modify the inner stream.
This is the first PR towards migrating to a new `$env.PWD` API that
returns potentially un-canonicalized paths. Refer to PR #12515 for
motivations.
## New API: `EngineState::cwd()`
The goal of the new API is to cover both parse-time and runtime use
case, and avoid unintentional misuse. It takes an `Option<Stack>` as
argument, which if supplied, will search for `$env.PWD` on the stack in
additional to the engine state. I think with this design, there's less
confusion over parse-time and runtime environments. If you have access
to a stack, just supply it; otherwise supply `None`.
## Deprecation of other PWD-related APIs
Other APIs are re-implemented using `EngineState::cwd()` and properly
documented. They're marked deprecated, but their behavior is unchanged.
Unused APIs are deleted, and code that accesses `$env.PWD` directly
without using an API is rewritten.
Deprecated APIs:
* `EngineState::current_work_dir()`
* `StateWorkingSet::get_cwd()`
* `env::current_dir()`
* `env::current_dir_str()`
* `env::current_dir_const()`
* `env::current_dir_str_const()`
Other changes:
* `EngineState::get_cwd()` (deleted)
* `StateWorkingSet::list_env()` (deleted)
* `repl::do_run_cmd()` (rewritten with `env::current_dir_str()`)
## `cd` and `pwd` now use logical paths by default
This pulls the changes from PR #12515. It's currently somewhat broken
because using non-canonicalized paths exposed a bug in our path
normalization logic (Issue #12602). Once that is fixed, this should
work.
## Future plans
This PR needs some tests. Which test helpers should I use, and where
should I put those tests?
I noticed that unquoted paths are expanded within `eval_filepath()` and
`eval_directory()` before they even reach the `cd` command. This means
every paths is expanded twice. Is this intended?
Once this PR lands, the plan is to review all usages of the deprecated
APIs and migrate them to `EngineState::cwd()`. In the meantime, these
usages are annotated with `#[allow(deprecated)]` to avoid breaking CI.
---------
Co-authored-by: Jakub Žádník <kubouch@gmail.com>
# Description
This helps to ensure data produced on a stream is immediately available
to the consumer of the stream. The BufWriter introduced for performance
reasons in 0.93 exposed the behavior that data messages wouldn't make it
to the other side until they filled the buffer in @cablehead's
[`nu_plugin_from_sse`](https://github.com/cablehead/nu_plugin_from_sse).
I had originally not flushed on every `Data` message because I figured
that it isn't really critical that the other side sees those messages
immediately, since they're not used for control and they are flushed
when waiting for acknowledgement or when the buffer is too full anyway.
Increasing the amount of data that can be sent with a single underlying
write increases performance, but this interferes with some plugins that
want to use streams in a more real-time way. In the future I would like
to make this configurable, maybe even per-command, so that a command can
decide what the priority is. But for now I think this is reasonable.
In the worst case, this decreases performance by about 40%, when sending
very small values (just numbers). But for larger values, this PR
actually increases performance by about 20%, because I've increased the
buffer size about 2x to 16,384 bytes. The previous value of 8,192 bytes
was too small to fit a full buffer coming from an external command, so
doubling it makes sense, and now a write of a buffer from an external
command can be done in exactly one write call, which I think makes
sense. I'm doing this at the same time because flushing each data
message would make it very likely that each individual data message from
an external stream would require exactly two writes rather than
approximately one (amortized).
Again, hopefully the tradeoff isn't too bad, and if it is I'll just make
it configurable.
# User-Facing Changes
- Performance of plugin streams will be a bit different
- Plugins that expect to send streams in real-time will work again
# Tests + Formatting
- 🟢 `toolkit fmt`
- 🟢 `toolkit clippy`
- 🟢 `toolkit test`
- 🟢 `toolkit test stdlib`
# Description
This breaks `nu-plugin` up into four crates:
- `nu-plugin-protocol`: just the type definitions for the protocol, no
I/O. If someone wanted to wire up something more bare metal, maybe for
async I/O, they could use this.
- `nu-plugin-core`: the shared stuff between engine/plugin. Less stable
interface.
- `nu-plugin-engine`: everything required for the engine to talk to
plugins. Less stable interface.
- `nu-plugin`: everything required for the plugin to talk to the engine,
what plugin developers use. Should be the most stable interface.
No changes are made to the interface exposed by `nu-plugin` - it should
all still be there. Re-exports from `nu-plugin-protocol` or
`nu-plugin-core` are used as required. Plugins shouldn't ever have to
use those crates directly.
This should be somewhat faster to compile as `nu-plugin-engine` and
`nu-plugin` can compile in parallel, and the engine doesn't need
`nu-plugin` and plugins don't need `nu-plugin-engine` (except for test
support), so that should reduce what needs to be compiled too.
The only significant change here other than splitting stuff up was to
break the `source` out of `PluginCustomValue` and create a new
`PluginCustomValueWithSource` type that contains that instead. One bonus
of that is we get rid of the option and it's now more type-safe, but it
also means that the logic for that stuff (actually running the plugin
for custom value ops) can live entirely within the `nu-plugin-engine`
crate.
# User-Facing Changes
- New crates.
- Added `local-socket` feature for `nu` to try to make it possible to
compile without that support if needed.
# Tests + Formatting
- 🟢 `toolkit fmt`
- 🟢 `toolkit clippy`
- 🟢 `toolkit test`
- 🟢 `toolkit test stdlib`
