Following up on #7180 with some feature cleanup:
- Move the `database` feature from `plugin` to `default`
- Rename the `database` feature to `sqlite`
- Remove `--features=extra` from a lot of scripts etc.
- No need to specify this, the `extra` feature is now the same as the
default feature set
- Remove the now-redundant 2nd Ubuntu test run
* Skeleton implementation
Lots and lots of TODOs
* Bootstrap simple CustomValue plugin support test
* Create nu_plugin_custom_value
* Skeleton for nu_plugin_custom_values
* Return a custom value from plugin
* Encode CustomValues from plugin calls as PluginResponse::PluginData
* Add new PluginCall variant CollapseCustomValue
* Handle CollapseCustomValue plugin calls
* Add CallInput::Data variant to CallInfo inputs
* Handle CallInfo with CallInput::Data plugin calls
* Send CallInput::Data if Value is PluginCustomValue from plugin calls
* Remove unnecessary boxing of plugins CallInfo
* Add fields needed to collapse PluginCustomValue to it
* Document PluginCustomValue and its purpose
* Impl collapsing using plugin calls in PluginCustomValue::to_base_value
* Implement proper typetag based deserialization for CoolCustomValue
* Test demonstrating that passing back a custom value to plugin works
* Added a failing test for describing plugin CustomValues
* Support describe for PluginCustomValues
- Add name to PluginResponse::PluginData
- Also turn it into a struct for clarity
- Add name to PluginCustomValue
- Return name field from PluginCustomValue
* Demonstrate that plugins can create and handle multiple CustomValues
* Add bincode to nu-plugin dependencies
This is for demonstration purposes, any schemaless binary seralization
format will work. I picked bincode since it's the most popular for Rust
but there are defintely better options out there for this usecase
* serde_json::Value -> Vec<u8>
* Update capnp schema for new CallInfo.input field
* Move call_input capnp serialization and deserialization into new file
* Deserialize Value's span from Value itself instead of passing call.head
I am not sure if this was correct and I am breaking it or if it was a
bug, I don't fully understand how nu creates and uses Spans. What should
reuse spans and what should recreate new ones?
But yeah it felt weird that the Value's Span was being ignored since in
the json serializer just uses the Value's Span
* Add call_info value round trip test
* Add capnp CallInput::Data serialization and deserialization support
* Add CallInfo::CollapseCustomValue to capnp schema
* Add capnp PluginCall::CollapseCustomValue serialization and deserialization support
* Add PluginResponse::PluginData to capnp schema
* Add capnp PluginResponse::PluginData serialization and deserialization support
* Switch plugins::custom_values tests to capnp
Both json and capnp would work now! Sadly I can't choose both at the
same time :(
* Add missing JsonSerializer round trip tests
* Handle plugin returning PluginData as a response to CollapseCustomValue
* Refactor plugin calling into a reusable function
Many less levels of indentation now!
* Export PluginData from nu_plugin
So plugins can create their very own serve_plugin with whatever
CustomValue behavior they may desire
* Error if CustomValue cannot be handled by Plugin
* Updated nu_with_plugins to handle new nushell
- Now it requires the plugin format and name to be passed in, because
we can't really guess the format
- It calls `register` with format and plugin path
- It creates a temporary folder and in it an empty temporary plugin.nu
so that the tests don't conflict with each other or with local copy of
plugin.nu
- Instead of passing the commands via stdin it passes them via the new
--commands command line argument
* Rename path to command for clarity
* Enable core_inc tests
Remove deprecated inc feature and replace with new plugin feature
* Update core_inc tests for new nu_with_plugins syntax
* Rework core_inc::can_only_apply_one
The new inc plugin doesn't error if passed more than one but instead
chooses the highest increment
* Gate all plugin tests behind feature = "plugin" instead of one by one
* Remove format!-like behavior from nu_with_plugins
nu_with_plugins had format!-like behavior where it would allow calls
such as this:
```rs
nu_with_plugins!(
cwd: "dir/",
"open {} | get {}",
"Cargo.toml",
"package.version"
)
```
And although nifty it seems to have never been used before and the same
can be achieved with a format! like so:
```rs
nu_with_plugins!(
cwd: "dir/",
format!("open {} | get {}", "Cargo.toml", "package.version")
)
```
So I am removing it to keep the complexity of the macro in check
* Add multi-plugin support to nu_with_plugins
Useful for testing interactions between plugins
* Alternative 1: run `cargo build` inside of tests
* Handle Windows by canonicalizing paths and add .exe
One VM install later and lots of learning about how command line
arguments work and here we are
We introduce the `plugin` nu sub command (`nu plugin`) with basic plugin
loading support. We can choose to load plugins from a directory. Originally
introduced to make integration tests faster (by not loading any plugins on startup at all)
but `nu plugin --load some_path ; test_pipeline_that_uses_plugins_just_loaded` does not see it.
Therefore, a `nu_with_plugins!` macro for tests was introduced on top of nu`s `--skip-plugins`
switch executable which is set to true when running the integration tests that use the `nu!` macro now..
* Working towards a PoC for wasm
* Move bson and sqlite to plugins
* proof of concept now working
* tests are green
* Add CI test for --no-default-features
* Fix some tests
* Fix clippy and windows build
* More fixes
* Fix the windows build
* Fix the windows test
* typo fixes
* Change signature to take in short-hand flags
* update help information
* Parse short-hand flags as their long counterparts
* lints
* Modified a couple tests to use shorthand flags
