nushell/crates/nu-source/src/pretty.rs

537 lines
14 KiB
Rust
Raw Normal View History

Restructure and streamline token expansion (#1123) Restructure and streamline token expansion The purpose of this commit is to streamline the token expansion code, by removing aspects of the code that are no longer relevant, removing pointless duplication, and eliminating the need to pass the same arguments to `expand_syntax`. The first big-picture change in this commit is that instead of a handful of `expand_` functions, which take a TokensIterator and ExpandContext, a smaller number of methods on the `TokensIterator` do the same job. The second big-picture change in this commit is fully eliminating the coloring traits, making coloring a responsibility of the base expansion implementations. This also means that the coloring tracer is merged into the expansion tracer, so you can follow a single expansion and see how the expansion process produced colored tokens. One side effect of this change is that the expander itself is marginally more error-correcting. The error correction works by switching from structured expansion to `BackoffColoringMode` when an unexpected token is found, which guarantees that all spans of the source are colored, but may not be the most optimal error recovery strategy. That said, because `BackoffColoringMode` only extends as far as a closing delimiter (`)`, `]`, `}`) or pipe (`|`), it does result in fairly granular correction strategy. The current code still produces an `Err` (plus a complete list of colored shapes) from the parsing process if any errors are encountered, but this could easily be addressed now that the underlying expansion is error-correcting. This commit also colors any spans that are syntax errors in red, and causes the parser to include some additional information about what tokens were expected at any given point where an error was encountered, so that completions and hinting could be more robust in the future. Co-authored-by: Jonathan Turner <jonathandturner@users.noreply.github.com> Co-authored-by: Andrés N. Robalino <andres@androbtech.com>
2020-01-21 23:45:03 +01:00
use crate::meta::Spanned;
use crate::term_colored::TermColored;
use crate::text::Text;
use derive_new::new;
use pretty::{BoxAllocator, DocAllocator};
use std::hash::Hash;
use termcolor::{Color, ColorSpec};
#[derive(Clone, Copy, Debug, Eq, Ord, PartialEq, PartialOrd, Hash)]
pub enum ShellStyle {
Delimiter,
Key,
Value,
Equals,
Kind,
Keyword,
Operator,
Variable,
Primitive,
Opaque,
Description,
Error,
}
impl From<ShellAnnotation> for ColorSpec {
fn from(ann: ShellAnnotation) -> ColorSpec {
match ann.style {
ShellStyle::Delimiter => ColorSpec::new()
.set_fg(Some(Color::White))
.set_intense(false)
.clone(),
ShellStyle::Key => ColorSpec::new()
2020-01-11 18:21:59 +01:00
.set_fg(Some(Color::Green))
.set_intense(true)
.clone(),
ShellStyle::Value => ColorSpec::new()
.set_fg(Some(Color::White))
.set_intense(true)
.clone(),
ShellStyle::Equals => ColorSpec::new()
2020-01-11 18:21:59 +01:00
.set_fg(Some(Color::Green))
.set_intense(true)
.clone(),
ShellStyle::Kind => ColorSpec::new().set_fg(Some(Color::Cyan)).clone(),
ShellStyle::Variable => ColorSpec::new()
.set_fg(Some(Color::Green))
.set_intense(true)
.clone(),
ShellStyle::Keyword => ColorSpec::new().set_fg(Some(Color::Magenta)).clone(),
ShellStyle::Operator => ColorSpec::new().set_fg(Some(Color::Yellow)).clone(),
ShellStyle::Primitive => ColorSpec::new()
.set_fg(Some(Color::Green))
.set_intense(true)
.clone(),
ShellStyle::Opaque => ColorSpec::new()
.set_fg(Some(Color::Yellow))
.set_intense(true)
.clone(),
ShellStyle::Description => ColorSpec::new()
2020-01-11 18:21:59 +01:00
.set_fg(Some(Color::Green))
.set_intense(true)
.clone(),
ShellStyle::Error => ColorSpec::new()
.set_fg(Some(Color::Red))
.set_intense(true)
.clone(),
}
}
}
#[derive(Clone, Copy, Eq, Ord, PartialEq, PartialOrd, Hash, new)]
pub struct ShellAnnotation {
style: ShellStyle,
}
impl std::fmt::Debug for ShellAnnotation {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
write!(f, "{:?}", self.style)
}
}
impl ShellAnnotation {
pub fn style(style: impl Into<ShellStyle>) -> ShellAnnotation {
ShellAnnotation {
style: style.into(),
}
}
}
pub type PrettyDebugDoc =
pretty::Doc<'static, pretty::BoxDoc<'static, ShellAnnotation>, ShellAnnotation>;
pub type PrettyDebugDocBuilder = pretty::DocBuilder<'static, pretty::BoxAllocator, ShellAnnotation>;
pub use self::DebugDocBuilder as b;
#[derive(Clone, new)]
pub struct DebugDocBuilder {
pub inner: PrettyDebugDocBuilder,
}
Restructure and streamline token expansion (#1123) Restructure and streamline token expansion The purpose of this commit is to streamline the token expansion code, by removing aspects of the code that are no longer relevant, removing pointless duplication, and eliminating the need to pass the same arguments to `expand_syntax`. The first big-picture change in this commit is that instead of a handful of `expand_` functions, which take a TokensIterator and ExpandContext, a smaller number of methods on the `TokensIterator` do the same job. The second big-picture change in this commit is fully eliminating the coloring traits, making coloring a responsibility of the base expansion implementations. This also means that the coloring tracer is merged into the expansion tracer, so you can follow a single expansion and see how the expansion process produced colored tokens. One side effect of this change is that the expander itself is marginally more error-correcting. The error correction works by switching from structured expansion to `BackoffColoringMode` when an unexpected token is found, which guarantees that all spans of the source are colored, but may not be the most optimal error recovery strategy. That said, because `BackoffColoringMode` only extends as far as a closing delimiter (`)`, `]`, `}`) or pipe (`|`), it does result in fairly granular correction strategy. The current code still produces an `Err` (plus a complete list of colored shapes) from the parsing process if any errors are encountered, but this could easily be addressed now that the underlying expansion is error-correcting. This commit also colors any spans that are syntax errors in red, and causes the parser to include some additional information about what tokens were expected at any given point where an error was encountered, so that completions and hinting could be more robust in the future. Co-authored-by: Jonathan Turner <jonathandturner@users.noreply.github.com> Co-authored-by: Andrés N. Robalino <andres@androbtech.com>
2020-01-21 23:45:03 +01:00
impl PrettyDebug for bool {
fn pretty(&self) -> DebugDocBuilder {
match self {
true => b::primitive("true"),
false => b::primitive("false"),
}
}
}
impl PrettyDebug for () {
fn pretty(&self) -> DebugDocBuilder {
b::primitive("nothing")
}
}
impl PrettyDebug for DebugDocBuilder {
fn pretty(&self) -> DebugDocBuilder {
self.clone()
}
}
impl std::ops::Add for DebugDocBuilder {
type Output = DebugDocBuilder;
fn add(self, rhs: DebugDocBuilder) -> DebugDocBuilder {
DebugDocBuilder::new(self.inner.append(rhs.inner))
}
}
impl DebugDocBuilder {
pub fn from_doc(doc: DebugDoc) -> DebugDocBuilder {
DebugDocBuilder {
inner: BoxAllocator.nil().append(doc),
}
}
pub fn blank() -> DebugDocBuilder {
BoxAllocator.nil().into()
}
pub fn delimiter(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Delimiter)
}
pub fn key(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Key)
}
pub fn value(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Value)
}
pub fn into_value(self) -> DebugDocBuilder {
self.inner
.annotate(ShellAnnotation::style(ShellStyle::Value))
.into()
}
pub fn equals() -> DebugDocBuilder {
DebugDocBuilder::styled("=", ShellStyle::Equals)
}
pub fn kind(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Kind)
}
pub fn into_kind(self) -> DebugDocBuilder {
self.inner
.annotate(ShellAnnotation::style(ShellStyle::Kind))
.into()
}
pub fn typed(kind: &str, value: DebugDocBuilder) -> DebugDocBuilder {
Restructure and streamline token expansion (#1123) Restructure and streamline token expansion The purpose of this commit is to streamline the token expansion code, by removing aspects of the code that are no longer relevant, removing pointless duplication, and eliminating the need to pass the same arguments to `expand_syntax`. The first big-picture change in this commit is that instead of a handful of `expand_` functions, which take a TokensIterator and ExpandContext, a smaller number of methods on the `TokensIterator` do the same job. The second big-picture change in this commit is fully eliminating the coloring traits, making coloring a responsibility of the base expansion implementations. This also means that the coloring tracer is merged into the expansion tracer, so you can follow a single expansion and see how the expansion process produced colored tokens. One side effect of this change is that the expander itself is marginally more error-correcting. The error correction works by switching from structured expansion to `BackoffColoringMode` when an unexpected token is found, which guarantees that all spans of the source are colored, but may not be the most optimal error recovery strategy. That said, because `BackoffColoringMode` only extends as far as a closing delimiter (`)`, `]`, `}`) or pipe (`|`), it does result in fairly granular correction strategy. The current code still produces an `Err` (plus a complete list of colored shapes) from the parsing process if any errors are encountered, but this could easily be addressed now that the underlying expansion is error-correcting. This commit also colors any spans that are syntax errors in red, and causes the parser to include some additional information about what tokens were expected at any given point where an error was encountered, so that completions and hinting could be more robust in the future. Co-authored-by: Jonathan Turner <jonathandturner@users.noreply.github.com> Co-authored-by: Andrés N. Robalino <andres@androbtech.com>
2020-01-21 23:45:03 +01:00
b::kind(kind) + b::delimit("[", value.group(), "]")
}
pub fn subtyped(
kind: &str,
subkind: impl std::fmt::Display,
value: DebugDocBuilder,
) -> DebugDocBuilder {
b::delimit(
"(",
(b::kind(kind) + b::delimit("[", b::kind(format!("{}", subkind)), "]")).group()
+ b::space()
+ value.group(),
")",
)
.group()
}
pub fn keyword(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Keyword)
}
pub fn var(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Variable)
}
pub fn operator(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Operator)
}
pub fn primitive(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(format!("{}", string), ShellStyle::Primitive)
}
pub fn opaque(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Opaque)
}
pub fn description(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Description)
}
pub fn error(string: impl std::fmt::Display) -> DebugDocBuilder {
DebugDocBuilder::styled(string, ShellStyle::Error)
}
pub fn delimit(start: &str, doc: DebugDocBuilder, end: &str) -> DebugDocBuilder {
DebugDocBuilder::delimiter(start) + doc + DebugDocBuilder::delimiter(end)
}
pub fn preceded(before: DebugDocBuilder, body: DebugDocBuilder) -> DebugDocBuilder {
if body.is_empty() {
body
} else {
before + body
}
}
pub fn surrounded_option(
before: Option<DebugDocBuilder>,
builder: Option<DebugDocBuilder>,
after: Option<DebugDocBuilder>,
) -> DebugDocBuilder {
match builder {
None => DebugDocBuilder::blank(),
Some(b) => b::option(before) + b + b::option(after),
}
}
pub fn preceded_option(
before: Option<DebugDocBuilder>,
builder: Option<DebugDocBuilder>,
) -> DebugDocBuilder {
DebugDocBuilder::surrounded_option(before, builder, None)
}
pub fn option(builder: Option<DebugDocBuilder>) -> DebugDocBuilder {
match builder {
None => DebugDocBuilder::blank(),
Some(b) => b,
}
}
pub fn space() -> DebugDocBuilder {
BoxAllocator.space().into()
}
pub fn newline() -> DebugDocBuilder {
BoxAllocator.newline().into()
}
pub fn is_empty(&self) -> bool {
matches!(&self.inner.1, pretty::Doc::Nil)
}
pub fn or(self, doc: DebugDocBuilder) -> DebugDocBuilder {
if self.is_empty() {
doc
} else {
self
}
}
pub fn group(self) -> DebugDocBuilder {
self.inner.group().into()
}
pub fn nest(self) -> DebugDocBuilder {
self.inner.nest(1).group().into()
}
pub fn intersperse_with_source<'a, T: PrettyDebugWithSource + 'a>(
list: impl IntoIterator<Item = &'a T>,
separator: DebugDocBuilder,
source: &str,
) -> DebugDocBuilder {
BoxAllocator
.intersperse(
list.into_iter().filter_map(|item| {
let item = item.pretty_debug(source);
if item.is_empty() {
None
} else {
Some(item)
}
}),
separator,
)
.into()
}
pub fn intersperse<T: PrettyDebug>(
list: impl IntoIterator<Item = T>,
separator: DebugDocBuilder,
) -> DebugDocBuilder {
BoxAllocator
.intersperse(
list.into_iter().filter_map(|item| {
let item = item.pretty();
if item.is_empty() {
None
} else {
Some(item)
}
}),
separator,
)
.into()
}
pub fn list(list: impl IntoIterator<Item = DebugDocBuilder>) -> DebugDocBuilder {
let mut result: DebugDocBuilder = BoxAllocator.nil().into();
for item in list {
result = result + item;
}
result
}
fn styled(string: impl std::fmt::Display, style: ShellStyle) -> DebugDocBuilder {
BoxAllocator
.text(string.to_string())
.annotate(ShellAnnotation::style(style))
.into()
}
}
impl std::ops::Deref for DebugDocBuilder {
type Target = PrettyDebugDocBuilder;
fn deref(&self) -> &Self::Target {
&self.inner
}
}
#[derive(Clone, Debug, Eq, Ord, PartialEq, PartialOrd, new)]
pub struct DebugDoc {
pub inner: PrettyDebugDoc,
}
Restructure and streamline token expansion (#1123) Restructure and streamline token expansion The purpose of this commit is to streamline the token expansion code, by removing aspects of the code that are no longer relevant, removing pointless duplication, and eliminating the need to pass the same arguments to `expand_syntax`. The first big-picture change in this commit is that instead of a handful of `expand_` functions, which take a TokensIterator and ExpandContext, a smaller number of methods on the `TokensIterator` do the same job. The second big-picture change in this commit is fully eliminating the coloring traits, making coloring a responsibility of the base expansion implementations. This also means that the coloring tracer is merged into the expansion tracer, so you can follow a single expansion and see how the expansion process produced colored tokens. One side effect of this change is that the expander itself is marginally more error-correcting. The error correction works by switching from structured expansion to `BackoffColoringMode` when an unexpected token is found, which guarantees that all spans of the source are colored, but may not be the most optimal error recovery strategy. That said, because `BackoffColoringMode` only extends as far as a closing delimiter (`)`, `]`, `}`) or pipe (`|`), it does result in fairly granular correction strategy. The current code still produces an `Err` (plus a complete list of colored shapes) from the parsing process if any errors are encountered, but this could easily be addressed now that the underlying expansion is error-correcting. This commit also colors any spans that are syntax errors in red, and causes the parser to include some additional information about what tokens were expected at any given point where an error was encountered, so that completions and hinting could be more robust in the future. Co-authored-by: Jonathan Turner <jonathandturner@users.noreply.github.com> Co-authored-by: Andrés N. Robalino <andres@androbtech.com>
2020-01-21 23:45:03 +01:00
#[derive(Debug, Copy, Clone)]
pub enum PrettyDebugRefineKind {
ContextFree,
WithContext,
}
pub trait PrettyDebugWithSource: Sized {
fn pretty_debug(&self, source: &str) -> DebugDocBuilder;
Restructure and streamline token expansion (#1123) Restructure and streamline token expansion The purpose of this commit is to streamline the token expansion code, by removing aspects of the code that are no longer relevant, removing pointless duplication, and eliminating the need to pass the same arguments to `expand_syntax`. The first big-picture change in this commit is that instead of a handful of `expand_` functions, which take a TokensIterator and ExpandContext, a smaller number of methods on the `TokensIterator` do the same job. The second big-picture change in this commit is fully eliminating the coloring traits, making coloring a responsibility of the base expansion implementations. This also means that the coloring tracer is merged into the expansion tracer, so you can follow a single expansion and see how the expansion process produced colored tokens. One side effect of this change is that the expander itself is marginally more error-correcting. The error correction works by switching from structured expansion to `BackoffColoringMode` when an unexpected token is found, which guarantees that all spans of the source are colored, but may not be the most optimal error recovery strategy. That said, because `BackoffColoringMode` only extends as far as a closing delimiter (`)`, `]`, `}`) or pipe (`|`), it does result in fairly granular correction strategy. The current code still produces an `Err` (plus a complete list of colored shapes) from the parsing process if any errors are encountered, but this could easily be addressed now that the underlying expansion is error-correcting. This commit also colors any spans that are syntax errors in red, and causes the parser to include some additional information about what tokens were expected at any given point where an error was encountered, so that completions and hinting could be more robust in the future. Co-authored-by: Jonathan Turner <jonathandturner@users.noreply.github.com> Co-authored-by: Andrés N. Robalino <andres@androbtech.com>
2020-01-21 23:45:03 +01:00
fn refined_pretty_debug(
&self,
_refine: PrettyDebugRefineKind,
source: &str,
) -> DebugDocBuilder {
self.pretty_debug(source)
}
// This is a transitional convenience method
fn debug(&self, source: impl Into<Text>) -> String
where
Self: Clone,
{
self.clone().debuggable(source).display()
}
fn debuggable(self, source: impl Into<Text>) -> DebuggableWithSource<Self> {
DebuggableWithSource {
inner: self,
source: source.into(),
}
}
}
Restructure and streamline token expansion (#1123) Restructure and streamline token expansion The purpose of this commit is to streamline the token expansion code, by removing aspects of the code that are no longer relevant, removing pointless duplication, and eliminating the need to pass the same arguments to `expand_syntax`. The first big-picture change in this commit is that instead of a handful of `expand_` functions, which take a TokensIterator and ExpandContext, a smaller number of methods on the `TokensIterator` do the same job. The second big-picture change in this commit is fully eliminating the coloring traits, making coloring a responsibility of the base expansion implementations. This also means that the coloring tracer is merged into the expansion tracer, so you can follow a single expansion and see how the expansion process produced colored tokens. One side effect of this change is that the expander itself is marginally more error-correcting. The error correction works by switching from structured expansion to `BackoffColoringMode` when an unexpected token is found, which guarantees that all spans of the source are colored, but may not be the most optimal error recovery strategy. That said, because `BackoffColoringMode` only extends as far as a closing delimiter (`)`, `]`, `}`) or pipe (`|`), it does result in fairly granular correction strategy. The current code still produces an `Err` (plus a complete list of colored shapes) from the parsing process if any errors are encountered, but this could easily be addressed now that the underlying expansion is error-correcting. This commit also colors any spans that are syntax errors in red, and causes the parser to include some additional information about what tokens were expected at any given point where an error was encountered, so that completions and hinting could be more robust in the future. Co-authored-by: Jonathan Turner <jonathandturner@users.noreply.github.com> Co-authored-by: Andrés N. Robalino <andres@androbtech.com>
2020-01-21 23:45:03 +01:00
impl<T: PrettyDebug> PrettyDebug for Spanned<T> {
fn pretty(&self) -> DebugDocBuilder {
self.item.pretty()
}
}
impl<T: PrettyDebug> PrettyDebugWithSource for T {
fn pretty_debug(&self, _source: &str) -> DebugDocBuilder {
self.pretty()
}
}
Restructure and streamline token expansion (#1123) Restructure and streamline token expansion The purpose of this commit is to streamline the token expansion code, by removing aspects of the code that are no longer relevant, removing pointless duplication, and eliminating the need to pass the same arguments to `expand_syntax`. The first big-picture change in this commit is that instead of a handful of `expand_` functions, which take a TokensIterator and ExpandContext, a smaller number of methods on the `TokensIterator` do the same job. The second big-picture change in this commit is fully eliminating the coloring traits, making coloring a responsibility of the base expansion implementations. This also means that the coloring tracer is merged into the expansion tracer, so you can follow a single expansion and see how the expansion process produced colored tokens. One side effect of this change is that the expander itself is marginally more error-correcting. The error correction works by switching from structured expansion to `BackoffColoringMode` when an unexpected token is found, which guarantees that all spans of the source are colored, but may not be the most optimal error recovery strategy. That said, because `BackoffColoringMode` only extends as far as a closing delimiter (`)`, `]`, `}`) or pipe (`|`), it does result in fairly granular correction strategy. The current code still produces an `Err` (plus a complete list of colored shapes) from the parsing process if any errors are encountered, but this could easily be addressed now that the underlying expansion is error-correcting. This commit also colors any spans that are syntax errors in red, and causes the parser to include some additional information about what tokens were expected at any given point where an error was encountered, so that completions and hinting could be more robust in the future. Co-authored-by: Jonathan Turner <jonathandturner@users.noreply.github.com> Co-authored-by: Andrés N. Robalino <andres@androbtech.com>
2020-01-21 23:45:03 +01:00
impl<T: PrettyDebugWithSource, E> PrettyDebugWithSource for Result<T, E> {
fn pretty_debug(&self, source: &str) -> DebugDocBuilder {
match self {
Err(_) => b::error("error"),
Ok(val) => val.pretty_debug(source),
}
}
}
pub struct DebuggableWithSource<T: PrettyDebugWithSource> {
inner: T,
source: Text,
}
impl<T> PrettyDebug for DebuggableWithSource<T>
where
T: PrettyDebugWithSource,
{
fn pretty(&self) -> DebugDocBuilder {
self.inner.pretty_debug(&self.source)
}
}
impl PrettyDebug for DebugDoc {
fn pretty(&self) -> DebugDocBuilder {
DebugDocBuilder::new(BoxAllocator.nil().append(self.inner.clone()))
}
}
pub trait PrettyDebug {
fn pretty(&self) -> DebugDocBuilder;
fn to_doc(&self) -> DebugDoc {
DebugDoc::new(self.pretty().into())
}
fn pretty_doc(&self) -> PrettyDebugDoc {
let builder = self.pretty();
builder.inner.into()
}
fn pretty_builder(&self) -> PrettyDebugDocBuilder {
let doc = self.pretty();
doc.inner
}
/// A convenience method that prints out the document without colors in
/// 70 columns. Generally, you should use plain_string or colored_string
/// if possible, but display() can be useful for trace lines and things
/// like that, where you don't have control over the terminal.
fn display(&self) -> String {
self.plain_string(70)
}
fn plain_string(&self, width: usize) -> String {
let doc = self.pretty_doc();
let mut buffer = termcolor::Buffer::no_color();
2020-01-02 06:24:41 +01:00
let _ = doc.render_raw(width, &mut TermColored::new(&mut buffer));
String::from_utf8_lossy(buffer.as_slice()).to_string()
}
fn colored_string(&self, width: usize) -> String {
let doc = self.pretty_doc();
let mut buffer = termcolor::Buffer::ansi();
2020-01-02 06:24:41 +01:00
let _ = doc.render_raw(width, &mut TermColored::new(&mut buffer));
String::from_utf8_lossy(buffer.as_slice()).to_string()
}
}
impl Into<DebugDocBuilder> for PrettyDebugDocBuilder {
fn into(self) -> DebugDocBuilder {
DebugDocBuilder { inner: self }
}
}
impl std::ops::Deref for DebugDoc {
type Target = PrettyDebugDoc;
fn deref(&self) -> &Self::Target {
&self.inner
}
}
impl From<DebugDoc> for PrettyDebugDoc {
fn from(input: DebugDoc) -> PrettyDebugDoc {
input.inner
}
}
impl Into<PrettyDebugDoc> for DebugDocBuilder {
fn into(self) -> PrettyDebugDoc {
self.inner.into()
}
}
fn hash_doc<H: std::hash::Hasher>(doc: &PrettyDebugDoc, state: &mut H) {
match doc {
pretty::Doc::Nil => 0u8.hash(state),
pretty::Doc::Append(a, b) => {
1u8.hash(state);
hash_doc(&*a, state);
hash_doc(&*b, state);
}
pretty::Doc::Group(a) => {
2u8.hash(state);
hash_doc(&*a, state);
}
pretty::Doc::Nest(a, b) => {
3u8.hash(state);
a.hash(state);
hash_doc(&*b, state);
}
pretty::Doc::Space => 4u8.hash(state),
pretty::Doc::Newline => 5u8.hash(state),
pretty::Doc::Text(t) => {
6u8.hash(state);
t.hash(state);
}
pretty::Doc::Annotated(a, b) => {
7u8.hash(state);
a.hash(state);
hash_doc(&*b, state);
}
}
}
#[allow(clippy::derive_hash_xor_eq)]
impl std::hash::Hash for DebugDoc {
fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
hash_doc(&self.inner, state);
}
}