nushell/crates/nu-protocol/src/value/column_path.rs
2020-10-15 17:25:17 -05:00

204 lines
6.6 KiB
Rust

use derive_new::new;
use getset::Getters;
use nu_source::{
b, span_for_spanned_list, DebugDocBuilder, HasFallibleSpan, PrettyDebug, Span, Spanned,
SpannedItem,
};
use num_bigint::BigInt;
use serde::{Deserialize, Serialize};
use crate::hir::{Expression, Literal, Member, SpannedExpression};
use nu_errors::ParseError;
/// A PathMember that has yet to be spanned so that it can be used in later processing
#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash, Serialize, Deserialize)]
pub enum UnspannedPathMember {
String(String),
Int(BigInt),
}
impl UnspannedPathMember {
/// Add the span information and get a full PathMember
pub fn into_path_member(self, span: impl Into<Span>) -> PathMember {
PathMember {
unspanned: self,
span: span.into(),
}
}
}
/// A basic piece of a ColumnPath, which describes the steps to take through a table to arrive a cell, row, or inner table
#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash, Serialize, Deserialize)]
pub struct PathMember {
pub unspanned: UnspannedPathMember,
pub span: Span,
}
impl PrettyDebug for &PathMember {
/// Gets the PathMember ready to be pretty-printed
fn pretty(&self) -> DebugDocBuilder {
match &self.unspanned {
UnspannedPathMember::String(string) => b::primitive(format!("{:?}", string)),
UnspannedPathMember::Int(int) => b::primitive(format!("{}", int)),
}
}
}
/// The fundamental path primitive to descrive how to navigate through a table to get to a sub-item. A path member can be either a word or a number. Words/strings are taken to mean
/// a column name, and numbers are the row number. Taken together they describe which column or row to narrow to in order to get data.
///
/// Rows must follow column names, they can't come first. eg) `foo.1` is valid where `1.foo` is not.
#[derive(
Debug, Hash, Serialize, Deserialize, Ord, PartialOrd, Eq, PartialEq, Getters, Clone, new,
)]
pub struct ColumnPath {
#[get = "pub"]
members: Vec<PathMember>,
}
impl ColumnPath {
/// Iterate over the members of the column path
pub fn iter(&self) -> impl Iterator<Item = &PathMember> {
self.members.iter()
}
/// Returns the last member and a slice of the remaining members
pub fn split_last(&self) -> Option<(&PathMember, &[PathMember])> {
self.members.split_last()
}
/// Returns the last member
pub fn last(&self) -> Option<&PathMember> {
self.iter().last()
}
pub fn build(text: &Spanned<String>) -> ColumnPath {
if let (
SpannedExpression {
expr: Expression::Literal(Literal::ColumnPath(path)),
span: _,
},
_,
) = parse(&text)
{
ColumnPath {
members: path.iter().map(|member| member.to_path_member()).collect(),
}
} else {
ColumnPath { members: vec![] }
}
}
}
impl PrettyDebug for ColumnPath {
/// Gets the ColumnPath ready to be pretty-printed
fn pretty(&self) -> DebugDocBuilder {
let members: Vec<DebugDocBuilder> =
self.members.iter().map(|member| member.pretty()).collect();
b::delimit(
"(",
b::description("path") + b::equals() + b::intersperse(members, b::space()),
")",
)
.nest()
}
}
impl HasFallibleSpan for ColumnPath {
/// Creates a span that will cover the column path, if possible
fn maybe_span(&self) -> Option<Span> {
if self.members.is_empty() {
None
} else {
Some(span_for_spanned_list(self.members.iter().map(|m| m.span)))
}
}
}
impl PathMember {
/// Create a string path member
pub fn string(string: impl Into<String>, span: impl Into<Span>) -> PathMember {
UnspannedPathMember::String(string.into()).into_path_member(span)
}
/// Create a numeric path member
pub fn int(int: impl Into<BigInt>, span: impl Into<Span>) -> PathMember {
UnspannedPathMember::Int(int.into()).into_path_member(span)
}
pub fn as_string(&self) -> String {
match &self.unspanned {
UnspannedPathMember::String(string) => string.clone(),
UnspannedPathMember::Int(int) => format!("{}", int),
}
}
}
fn parse(raw_column_path: &Spanned<String>) -> (SpannedExpression, Option<ParseError>) {
let mut delimiter = '.';
let mut inside_delimiter = false;
let mut output = vec![];
let mut current_part = String::new();
let mut start_index = 0;
let mut last_index = 0;
for (idx, c) in raw_column_path.item.char_indices() {
last_index = idx;
if inside_delimiter {
if c == delimiter {
inside_delimiter = false;
}
} else if c == '\'' || c == '"' || c == '`' {
inside_delimiter = true;
delimiter = c;
} else if c == '.' {
let part_span = Span::new(
raw_column_path.span.start() + start_index,
raw_column_path.span.start() + idx,
);
if let Ok(row_number) = current_part.parse::<u64>() {
output.push(Member::Int(BigInt::from(row_number), part_span));
} else {
let trimmed = trim_quotes(&current_part);
output.push(Member::Bare(trimmed.clone().spanned(part_span)));
}
current_part.clear();
// Note: I believe this is safe because of the delimiter we're using, but if we get fancy with
// unicode we'll need to change this
start_index = idx + '.'.len_utf8();
continue;
}
current_part.push(c);
}
if !current_part.is_empty() {
let part_span = Span::new(
raw_column_path.span.start() + start_index,
raw_column_path.span.start() + last_index + 1,
);
if let Ok(row_number) = current_part.parse::<u64>() {
output.push(Member::Int(BigInt::from(row_number), part_span));
} else {
let current_part = trim_quotes(&current_part);
output.push(Member::Bare(current_part.spanned(part_span)));
}
}
(
SpannedExpression::new(Expression::simple_column_path(output), raw_column_path.span),
None,
)
}
fn trim_quotes(input: &str) -> String {
let mut chars = input.chars();
match (chars.next(), chars.next_back()) {
(Some('\''), Some('\'')) => chars.collect(),
(Some('"'), Some('"')) => chars.collect(),
(Some('`'), Some('`')) => chars.collect(),
_ => input.to_string(),
}
}