Enum nu_protocol::PipelineData

pub enum PipelineData {
    Value(Value, Option<PipelineMetadata>),
    ListStream(ListStream, Option<PipelineMetadata>),
    ExternalStream {
        stdout: Option<RawStream>,
        stderr: Option<RawStream>,
        exit_code: Option<ListStream>,
        span: Span,
        metadata: Option<PipelineMetadata>,
        trim_end_newline: bool,
    },
    Empty,
}

The foundational abstraction for input and output to commands

This represents either a single Value or a stream of values coming into the command or leaving a command.

A note on implementation:

We’ve tried a few variations of this structure. Listing these below so we have a record.

• We tried always assuming a stream in Nushell. This was a great 80% solution, but it had some rough edges. Namely, how do you know the difference between a single string and a list of one string. How do you know when to flatten the data given to you from a data source into the stream or to keep it as an unflattened list?

• We tried putting the stream into Value. This had some interesting properties as now commands “just worked on values”, but lead to a few unfortunate issues.

The first is that you can’t easily clone Values in a way that felt largely immutable. For example, if you cloned a Value which contained a stream, and in one variable drained some part of it, then the second variable would see different values based on what you did to the first.

To make this kind of mutation thread-safe, we would have had to produce a lock for the stream, which in practice would have meant always locking the stream before reading from it. But more fundamentally, it felt wrong in practice that observation of a value at runtime could affect other values which happen to alias the same stream. By separating these, we don’t have this effect. Instead, variables could get concrete list values rather than streams, and be able to view them without non-local effects.

• A balance of the two approaches is what we’ve landed on: Values are thread-safe to pass, and we can stream them into any sources. Streams are still available to model the infinite streams approach of original Nushell.

Variants

Value(Value, Option<PipelineMetadata>)

ListStream(ListStream, Option<PipelineMetadata>)

ExternalStream

Fields

stdout: Option<RawStream>

stderr: Option<RawStream>

exit_code: Option<ListStream>

span: Span

metadata: Option<PipelineMetadata>

trim_end_newline: bool

Empty

Implementations

impl PipelineData

pub fn new_with_metadata( metadata: Option<PipelineMetadata>, span: Span ) -> PipelineData

pub fn new_external_stream_with_only_exit_code(exit_code: i64) -> PipelineData

create a PipelineData::ExternalStream with proper exit_code

It’s useful to break running without raising error at user level.

pub fn empty() -> PipelineData

pub fn metadata(&self) -> Option<PipelineMetadata>

pub fn set_metadata(self, metadata: Option<PipelineMetadata>) -> Self

pub fn is_nothing(&self) -> bool

pub fn span(&self) -> Option<Span>

PipelineData doesn’t always have a Span, but we can try!

pub fn into_value(self, span: Span) -> Value

pub fn write_to_out_dests( self, engine_state: &EngineState, stack: &mut Stack ) -> Result<PipelineData, ShellError>

Writes all values or redirects all output to the current OutDests in stack.

For OutDest::Pipe and OutDest::Capture, this will return the PipelineData as is without consuming input and without writing anything.

For the other OutDests, the given PipelineData will be completely consumed and PipelineData::Empty will be returned.

pub fn drain(self) -> Result<(), ShellError>

pub fn drain_with_exit_code(self) -> Result<i64, ShellError>

pub fn into_iter_strict( self, span: Span ) -> Result<PipelineIterator, ShellError>

Try convert from self into iterator

It returns Err if the self cannot be converted to an iterator.

pub fn into_interruptible_iter( self, ctrlc: Option<Arc<AtomicBool>> ) -> PipelineIterator

pub fn collect_string( self, separator: &str, config: &Config ) -> Result<String, ShellError>

pub fn collect_string_strict( self, span: Span ) -> Result<(String, Span, Option<PipelineMetadata>), ShellError>

Retrieves string from pipeline data.

As opposed to collect_string this raises error rather than converting non-string values. The span will be used if ListStream is encountered since it doesn’t carry a span.

pub fn follow_cell_path( self, cell_path: &[PathMember], head: Span, insensitive: bool ) -> Result<Value, ShellError>

pub fn upsert_cell_path( &mut self, cell_path: &[PathMember], callback: Box<dyn FnOnce(&Value) -> Value>, head: Span ) -> Result<(), ShellError>

pub fn map<F>( self, f: F, ctrlc: Option<Arc<AtomicBool>> ) -> Result<PipelineData, ShellError>

where Self: Sized, F: FnMut(Value) -> Value + 'static + Send,

Simplified mapper to help with simple values also. For full iterator support use .into_iter() instead

pub fn flat_map<U, F>( self, f: F, ctrlc: Option<Arc<AtomicBool>> ) -> Result<PipelineData, ShellError>

where Self: Sized, U: IntoIterator<Item = Value> + 'static, <U as IntoIterator>::IntoIter: 'static + Send, F: FnMut(Value) -> U + 'static + Send,

Simplified flatmapper. For full iterator support use .into_iter() instead

pub fn filter<F>( self, f: F, ctrlc: Option<Arc<AtomicBool>> ) -> Result<PipelineData, ShellError>

where Self: Sized, F: FnMut(&Value) -> bool + 'static + Send,

pub fn check_external_failed(self) -> (Self, bool)

Try to catch the external stream exit status and detect if it failed.

This is useful for external commands with semicolon, we can detect errors early to avoid commands after the semicolon running.

Returns self and a flag that indicates if the external stream run failed. If self is not PipelineData::ExternalStream, the flag will be false.

Currently this will consume an external stream to completion.

pub fn try_expand_range(self) -> Result<PipelineData, ShellError>

Try to convert Value from Value::Range to Value::List. This is useful to expand Value::Range into array notation, specifically when converting to json or to nuon. 1..3 | to XX -> [1,2,3]

pub fn print( self, engine_state: &EngineState, stack: &mut Stack, no_newline: bool, to_stderr: bool ) -> Result<i64, ShellError>

Consume and print self data immediately.

no_newline controls if we need to attach newline character to output. to_stderr controls if data is output to stderr, when the value is false, the data is output to stdout.

pub fn print_not_formatted( self, engine_state: &EngineState, no_newline: bool, to_stderr: bool ) -> Result<i64, ShellError>

Consume and print self data immediately.

Unlike [.print()] does not call table to format data and just prints it one element on a line

• no_newline controls if we need to attach newline character to output.
• to_stderr controls if data is output to stderr, when the value is false, the data is output to stdout.

Trait Implementations

impl Debug for PipelineData

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl IntoIterator for PipelineData

type Item = Value

The type of the elements being iterated over.

type IntoIter = PipelineIterator

Which kind of iterator are we turning this into?

fn into_iter(self) -> Self::IntoIter

Creates an iterator from a value.