diff --git a/.editorconfig b/.editorconfig index c6c8b3621..f6d83b7c7 100644 --- a/.editorconfig +++ b/.editorconfig @@ -7,3 +7,11 @@ end_of_line = lf charset = utf-8 trim_trailing_whitespace = true insert_final_newline = true + +[*.rs] +indent_style = space +indent_size = 4 +end_of_line = lf +charset = utf-8 +trim_trailing_whitespace = true +insert_final_newline = true diff --git a/oxide/crates/core/benches/parse_candidates.rs b/oxide/crates/core/benches/parse_candidates.rs index 55c919471..aacc1469e 100644 --- a/oxide/crates/core/benches/parse_candidates.rs +++ b/oxide/crates/core/benches/parse_candidates.rs @@ -31,6 +31,18 @@ pub fn criterion_benchmark(c: &mut Criterion) { c.bench_function("parse_candidate_strings (real world)", |b| { b.iter(|| parse(include_bytes!("./fixtures/template-499.html"))) }); + + let mut group = c.benchmark_group("sample-size-example"); + group.sample_size(10); + + group.bench_function("parse_candidate_strings (fast space skipping)", |b| { + let count = 10_000; + let crazy1 = format!("{}underline", " ".repeat(count)); + let crazy2 = crazy1.repeat(count); + let crazy3 = crazy2.as_bytes(); + + b.iter(|| parse(black_box(crazy3))) + }); } criterion_group!(benches, criterion_benchmark); diff --git a/oxide/crates/core/src/cursor.rs b/oxide/crates/core/src/cursor.rs new file mode 100644 index 000000000..0c422c6ff --- /dev/null +++ b/oxide/crates/core/src/cursor.rs @@ -0,0 +1,159 @@ +use std::{ascii::escape_default, fmt::Display}; + +#[derive(Debug, Clone)] +pub struct Cursor<'a> { + // The input we're scanning + pub input: &'a [u8], + + // The location of the cursor in the input + pub pos: usize, + + /// Is the cursor at the start of the input + pub at_start: bool, + + /// Is the cursor at the end of the input + pub at_end: bool, + + /// The previously consumed character + /// If `at_start` is true, this will be NUL + pub prev: u8, + + /// The current character + pub curr: u8, + + /// The upcoming character (if any) + /// If `at_end` is true, this will be NUL + pub next: u8, +} + +impl<'a> Cursor<'a> { + pub fn new(input: &'a [u8]) -> Self { + let mut cursor = Self { + input, + pos: 0, + at_start: true, + at_end: false, + prev: 0x00, + curr: 0x00, + next: 0x00, + }; + cursor.move_to(0); + cursor + } + + pub fn rewind_by(&mut self, amount: usize) { + self.move_to(self.pos.saturating_sub(amount)); + } + + pub fn advance_by(&mut self, amount: usize) { + self.move_to(self.pos.saturating_add(amount)); + } + + pub fn move_to(&mut self, pos: usize) { + let len = self.input.len(); + let pos = pos.clamp(0, len); + + self.pos = pos; + self.at_start = pos == 0; + self.at_end = pos + 1 >= len; + + self.prev = if pos > 0 { self.input[pos - 1] } else { 0x00 }; + self.curr = if pos < len { self.input[pos] } else { 0x00 }; + self.next = if pos + 1 < len { + self.input[pos + 1] + } else { + 0x00 + }; + } +} + +impl<'a> Display for Cursor<'a> { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + let len = self.input.len().to_string(); + + let pos = format!("{: >len_count$}", self.pos, len_count = len.len()); + write!(f, "{}/{} ", pos, len)?; + + if self.at_start { + write!(f, "S ")?; + } else if self.at_end { + write!(f, "E ")?; + } else { + write!(f, "M ")?; + } + + fn to_str(c: u8) -> String { + if c == 0x00 { + "NUL".into() + } else { + format!("{:?}", escape_default(c).to_string()) + } + } + + write!( + f, + "[{} {} {}]", + to_str(self.prev), + to_str(self.curr), + to_str(self.next) + ) + } +} + +#[cfg(test)] +mod test { + use super::*; + + #[test] + fn test_cursor() { + let mut cursor = Cursor::new(b"hello world"); + assert_eq!(cursor.pos, 0); + assert!(cursor.at_start); + assert!(!cursor.at_end); + assert_eq!(cursor.prev, 0x00); + assert_eq!(cursor.curr, b'h'); + assert_eq!(cursor.next, b'e'); + + cursor.advance_by(1); + assert_eq!(cursor.pos, 1); + assert!(!cursor.at_start); + assert!(!cursor.at_end); + assert_eq!(cursor.prev, b'h'); + assert_eq!(cursor.curr, b'e'); + assert_eq!(cursor.next, b'l'); + + // Advancing too far should stop at the end + cursor.advance_by(10); + assert_eq!(cursor.pos, 11); + assert!(!cursor.at_start); + assert!(cursor.at_end); + assert_eq!(cursor.prev, b'd'); + assert_eq!(cursor.curr, 0x00); + assert_eq!(cursor.next, 0x00); + + // Can't advance past the end + cursor.advance_by(1); + assert_eq!(cursor.pos, 11); + assert!(!cursor.at_start); + assert!(cursor.at_end); + assert_eq!(cursor.prev, b'd'); + assert_eq!(cursor.curr, 0x00); + assert_eq!(cursor.next, 0x00); + + cursor.rewind_by(1); + assert_eq!(cursor.pos, 10); + assert!(!cursor.at_start); + assert!(cursor.at_end); + assert_eq!(cursor.prev, b'l'); + assert_eq!(cursor.curr, b'd'); + assert_eq!(cursor.next, 0x00); + + cursor.rewind_by(10); + assert_eq!(cursor.pos, 0); + assert!(cursor.at_start); + assert!(!cursor.at_end); + assert_eq!(cursor.prev, 0x00); + assert_eq!(cursor.curr, b'h'); + assert_eq!(cursor.next, b'e'); + } +} diff --git a/oxide/crates/core/src/fast_skip.rs b/oxide/crates/core/src/fast_skip.rs new file mode 100644 index 000000000..488e61bd6 --- /dev/null +++ b/oxide/crates/core/src/fast_skip.rs @@ -0,0 +1,89 @@ +use crate::cursor::Cursor; + +const STRIDE: usize = 16; +type Mask = [bool; STRIDE]; + +#[inline(always)] +pub fn fast_skip(cursor: &Cursor) -> Option { + // If we don't have enough bytes left to check then bail early + if cursor.pos + STRIDE >= cursor.input.len() { + return None; + } + + if !cursor.curr.is_ascii_whitespace() { + return None; + } + + let mut offset = 1; + + // SAFETY: We've already checked (indirectly) that this index is valid + let remaining = unsafe { cursor.input.get_unchecked(cursor.pos..) }; + + // NOTE: This loop uses primitives designed to be auto-vectorized + // Do not change this loop without benchmarking the results + // And checking the generated assembly using godbolt.org + for (i, chunk) in remaining.chunks_exact(STRIDE).enumerate() { + let value = load(chunk); + let is_whitespace = is_ascii_whitespace(value); + let is_all_whitespace = all_true(is_whitespace); + + if is_all_whitespace { + offset = (i + 1) * STRIDE; + } else { + break; + } + } + + Some(cursor.pos + offset) +} + +#[inline(always)] +fn load(input: &[u8]) -> [u8; STRIDE] { + let mut value = [0u8; STRIDE]; + value.copy_from_slice(input); + value +} + +#[inline(always)] +fn eq(input: [u8; STRIDE], val: u8) -> Mask { + let mut res = [false; STRIDE]; + for n in 0..STRIDE { + res[n] = input[n] == val + } + res +} + +#[inline(always)] +fn or(a: [bool; STRIDE], b: [bool; STRIDE]) -> [bool; STRIDE] { + let mut res = [false; STRIDE]; + for n in 0..STRIDE { + res[n] = a[n] | b[n]; + } + res +} + +#[inline(always)] +fn all_true(a: [bool; STRIDE]) -> bool { + let mut res = true; + for item in a.iter().take(STRIDE) { + res &= item; + } + res +} + +#[inline(always)] +fn is_ascii_whitespace(value: [u8; STRIDE]) -> [bool; STRIDE] { + let whitespace_1 = eq(value, b'\t'); + let whitespace_2 = eq(value, b'\n'); + let whitespace_3 = eq(value, b'\x0C'); + let whitespace_4 = eq(value, b'\r'); + let whitespace_5 = eq(value, b' '); + + or( + or( + or(or(whitespace_1, whitespace_2), whitespace_3), + whitespace_4, + ), + whitespace_5, + ) +} diff --git a/oxide/crates/core/src/lib.rs b/oxide/crates/core/src/lib.rs index 858e17a2d..9596c6855 100644 --- a/oxide/crates/core/src/lib.rs +++ b/oxide/crates/core/src/lib.rs @@ -9,6 +9,8 @@ use tracing::event; use walkdir::WalkDir; pub mod candidate; +pub mod cursor; +pub mod fast_skip; pub mod glob; pub mod location; pub mod modifier; diff --git a/oxide/crates/core/src/parser.rs b/oxide/crates/core/src/parser.rs index acee57022..92740847b 100644 --- a/oxide/crates/core/src/parser.rs +++ b/oxide/crates/core/src/parser.rs @@ -2,6 +2,39 @@ use bstr::ByteSlice; use fxhash::FxHashSet; use tracing::trace; +use crate::{cursor::Cursor, fast_skip::fast_skip}; + +#[derive(Debug, PartialEq, Eq, Clone)] +pub enum ParseAction<'a> { + Consume, + Skip, + RestartAt(usize), + + SingleCandidate(&'a [u8]), + MultipleCandidates(Vec<&'a [u8]>), + Done, +} + +#[derive(Debug, PartialEq, Eq, Clone)] +pub enum Bracketing<'a> { + Included(&'a [u8]), + Wrapped(&'a [u8]), + None, +} + +#[derive(Debug, PartialEq, Eq, Clone, Copy)] +pub struct SplitCandidate<'a> { + variant: &'a [u8], + utility: &'a [u8], +} + +#[derive(Debug, PartialEq, Eq, Clone, Copy)] +pub enum ValidationResult { + Invalid, + Valid, + Restart, +} + #[derive(Default)] pub struct ExtractorOptions { pub preserve_spaces_in_arbitrary: bool, @@ -11,9 +44,7 @@ pub struct Extractor<'a> { opts: ExtractorOptions, input: &'a [u8], - pos: usize, - - prev: u8, + cursor: Cursor<'a>, idx_start: usize, idx_end: usize, @@ -24,24 +55,24 @@ pub struct Extractor<'a> { in_candidate: bool, in_escape: bool, + discard_next: bool, + quote_stack: Vec, bracket_stack: Vec, - // buffer: [Option<&'a [u8]>; 8], } impl<'a> Extractor<'a> { pub fn all(input: &'a [u8], opts: ExtractorOptions) -> Vec<&'a [u8]> { - Self::new(input, opts).collect() + Self::new(input, opts).flatten().collect() } pub fn unique(input: &'a [u8], opts: ExtractorOptions) -> FxHashSet<&'a [u8]> { let mut candidates: FxHashSet<&[u8]> = Default::default(); candidates.reserve(100); - candidates.extend(Self::new(input, opts)); + candidates.extend(Self::new(input, opts).flatten()); candidates } - #[cfg(test)] pub fn unique_ord(input: &'a [u8], opts: ExtractorOptions) -> Vec<&'a [u8]> { // This is an inefficient way to get an ordered, unique // list as a Vec but it is only meant for testing. @@ -59,8 +90,7 @@ impl<'a> Extractor<'a> { Self { opts, input, - pos: 0, - prev: 0, + cursor: Cursor::new(input), idx_start: 0, idx_end: 0, @@ -70,10 +100,11 @@ impl<'a> Extractor<'a> { in_candidate: false, in_escape: false, + discard_next: false, + idx_last: input.len(), quote_stack: Vec::with_capacity(8), bracket_stack: Vec::with_capacity(8), - // buffer: [None; 8], } } } @@ -86,126 +117,212 @@ impl<'a> Extractor<'a> { } #[inline(always)] - fn get_current_candidate(&mut self) -> Option<&'a [u8]> { - let candidate = &self.input[self.idx_start..=self.idx_end]; + fn get_current_candidate(&mut self) -> ParseAction<'a> { + if self.discard_next { + return ParseAction::Skip; + } - if Extractor::is_valid_candidate_string(candidate) { - Some(candidate) - } else { - None + let mut candidate = &self.input[self.idx_start..=self.idx_end]; + + while !candidate.is_empty() { + match Extractor::is_valid_candidate_string(candidate) { + ValidationResult::Valid => return ParseAction::SingleCandidate(candidate), + ValidationResult::Restart => return ParseAction::RestartAt(self.idx_start + 1), + _ => {} + } + + match candidate.split_last() { + // At this point the candidate is technically invalid, however it can be that it + // has a few dangling characters attached to it. For example, think about a + // JavaScript object: + // + // ```js + // { underline: true } + // ``` + // + // The candidate at this point will be `underline:`, which is invalid. However, we + // can assume in this case that the `:` should not be there, and therefore we can + // try to slice it off and retry the validation. + Some((b':' | b'/' | b'.', head)) => { + candidate = head; + } + + // It could also be that we have the candidate is nested inside of bracket or quote + // pairs. In this case we want to retrieve the inner part and try to validate that + // inner part instead. For example, in a JavaScript array: + // + // ```js + // let myClasses = ["underline"] + // ``` + // + // The `underline` is nested inside of quotes and in square brackets. Let's try to + // get the inner part and validate that instead. + _ => match Self::slice_surrounding(candidate) { + Some(shorter) if shorter != candidate => { + candidate = shorter; + } + _ => break, + }, + } + } + + ParseAction::Consume + } + + #[inline(always)] + fn split_candidate(candidate: &'a [u8]) -> SplitCandidate { + let mut brackets = 0; + let mut idx_end = 0; + + for (n, c) in candidate.iter().enumerate() { + match c { + b'[' => brackets += 1, + b']' if brackets > 0 => brackets -= 1, + b':' if brackets == 0 => idx_end = n + 1, + _ => {} + } + } + + SplitCandidate { + variant: &candidate[0..idx_end], + utility: &candidate[idx_end..], } } #[inline(always)] - fn is_valid_candidate_string(candidate: &'a [u8]) -> bool { - let original_candidate = &candidate; - let mut offset = 0; + fn contains_in_constrained(candidate: &'a [u8], bytes: Vec) -> bool { + let mut brackets = 0; - // Some special cases that we can ignore while testing the validations. - if candidate.starts_with(b"!-") { + for c in candidate { + match c { + b'[' => brackets += 1, + b']' if brackets > 0 => brackets -= 1, + _ if brackets == 0 && bytes.contains(c) => return true, + _ => {} + } + } + + false + } + + #[inline(always)] + fn is_valid_candidate_string(candidate: &'a [u8]) -> ValidationResult { + let split_candidate = Extractor::split_candidate(candidate); + + let mut offset = 0; + let utility = &split_candidate.utility; + let original_utility = &utility; + + // Some special cases that we can ignore while validating + if utility.starts_with(b"!-") { offset += 2; - } else if candidate.starts_with(b"!") || candidate.starts_with(b"-") { + } else if utility.starts_with(b"!") || utility.starts_with(b"-") { offset += 1; } + // These are allowed in arbitrary values and in variants but nowhere else + if Extractor::contains_in_constrained(utility, vec![b'<', b'>']) { + return ValidationResult::Restart; + } + // Pluck out the part that we are interested in. - let candidate = &candidate[offset..]; + let utility = &utility[offset..]; // Validations // We should have _something_ - if candidate.is_empty() { - return false; + if utility.is_empty() { + return ValidationResult::Invalid; } // = b'0' && candidate[0] <= b'9' && !candidate.contains(&b':') { - return false; + if utility[0] >= b'0' && utility[0] <= b'9' && !utility.contains(&b':') { + return ValidationResult::Invalid; } // In case of an arbitrary property, we should have at least this structure: [a:b] - if candidate.starts_with(b"[") && candidate.ends_with(b"]") { + if utility.starts_with(b"[") && utility.ends_with(b"]") { // [a:b] is at least 5 characters long - if candidate.len() < 5 { - return false; + if utility.len() < 5 { + return ValidationResult::Invalid; } // Should contain a `:` - if !candidate.contains(&b':') { - return false; + if !utility.contains(&b':') { + return ValidationResult::Invalid; } // Now that we validated that the candidate is technically fine, let's ensure that it // doesn't start with a `-` because that would make it invalid for arbitrary properties. - if original_candidate.starts_with(b"-") || original_candidate.starts_with(b"!-") { - return false; + if original_utility.starts_with(b"-") || original_utility.starts_with(b"!-") { + return ValidationResult::Invalid; } // The ':` must be preceded by a-Z0-9 because it represents a property name. - let colon = candidate.find(":").unwrap(); + let colon = utility.find(":").unwrap(); - if !candidate + if !utility .chars() .nth(colon - 1) .map_or_else(|| false, |c| c.is_ascii_alphanumeric()) { - return false; + return ValidationResult::Invalid; } } // In case of an arbitrary property, we should not have a modifier of any kind - if candidate.starts_with(b"[") { + if utility.starts_with(b"[") { if let Some(first_non_escaped_closing_bracket_idx) = - candidate.char_indices().position(|(s, _, other)| { + utility.char_indices().position(|(s, _, other)| { other == ']' - && candidate + && utility .get((s - 1)..s) .map(|c| !c.eq(b"\\")) .unwrap_or(false) }) { - if let Some(next) = candidate + if let Some(next) = utility .chars() .nth(first_non_escaped_closing_bracket_idx + 1) { // `/` is the indicator for a modifier, this is not allowed after arbitrary // properties if next == '/' { - return false; + return ValidationResult::Invalid; } } } } - true + ValidationResult::Valid } #[inline(always)] - fn parse_escaped(&mut self) -> bool { + fn parse_escaped(&mut self) -> ParseAction<'a> { // If this character is escaped, we don't care about it. // It gets consumed. trace!("Escape::Consume"); self.in_escape = false; - true + ParseAction::Consume } #[inline(always)] - fn parse_arbitrary(&mut self, curr: u8, pos: usize) -> bool { + fn parse_arbitrary(&mut self) -> ParseAction<'a> { // In this we could technically use memchr 6 times (then looped) to find the indexes / bounds of arbitrary valuesq if self.in_escape { return self.parse_escaped(); } - match curr { + match self.cursor.curr { b'\\' => { // The `\` character is used to escape characters in arbitrary content _and_ to prevent the starting of arbitrary content trace!("Arbitrary::Escape"); @@ -213,7 +330,7 @@ impl<'a> Extractor<'a> { } // Make sure the brackets are balanced - b'[' => self.bracket_stack.push(curr), + b'[' => self.bracket_stack.push(self.cursor.curr), b']' => match self.bracket_stack.last() { // We've ended a nested bracket Some(&last_bracket) if last_bracket == b'[' => { @@ -222,12 +339,16 @@ impl<'a> Extractor<'a> { // This is the last bracket meaning the end of arbitrary content _ if !self.in_quotes() => { + if matches!(self.cursor.next, b'a'..=b'z' | b'A'..=b'Z' | b'0'..=b'9') { + return ParseAction::Consume; + } + trace!("Arbitrary::End\t"); self.in_arbitrary = false; - if pos - self.idx_arbitrary_start == 1 { + if self.cursor.pos - self.idx_arbitrary_start == 1 { // We have an empty arbitrary value, which is not allowed - return false; + return ParseAction::Skip; } } @@ -239,19 +360,22 @@ impl<'a> Extractor<'a> { // These can "escape" the arbitrary value mode // switching of `[` and `]` characters b'"' | b'\'' | b'`' => match self.quote_stack.last() { - Some(&last_quote) if last_quote == curr => { + Some(&last_quote) if last_quote == self.cursor.curr => { trace!("Quote::End\t"); self.quote_stack.pop(); } _ => { trace!("Quote::Start\t"); - self.quote_stack.push(curr); + self.quote_stack.push(self.cursor.curr); } }, b' ' if !self.opts.preserve_spaces_in_arbitrary => { trace!("Arbitrary::SkipAndEndEarly\t"); - return false; + + // Restart the parser ahead of the arbitrary value + // It may pick up more candidates + return ParseAction::RestartAt(self.idx_arbitrary_start + 1); } // Arbitrary values allow any character inside them @@ -262,51 +386,55 @@ impl<'a> Extractor<'a> { } } - true + ParseAction::Consume } #[inline(always)] - fn parse_start(&mut self, curr: u8, pos: usize) -> bool { - match curr { + fn parse_start(&mut self) -> ParseAction<'a> { + match self.cursor.curr { // Enter arbitrary value mode b'[' => { trace!("Arbitrary::Start\t"); self.in_arbitrary = true; - self.idx_arbitrary_start = pos; + self.idx_arbitrary_start = self.cursor.pos; - true + ParseAction::Consume } // Allowed first characters. - b'@' | b'!' | b'-' | b'<' | b'0'..=b'9' | b'a'..=b'z' | b'A'..=b'Z' => { + b'@' | b'!' | b'-' | b'<' | b'>' | b'0'..=b'9' | b'a'..=b'z' | b'A'..=b'Z' => { // TODO: A bunch of characters that we currently support but maybe we only want it behind // a flag. E.g.: '' | '$' | '^' | '_' + // | '$' | '^' | '_' + + // When the new candidate is preceeded by a `:`, then we want to keep parsing, but + // throw away the full candidate because it can not be a valid candidate at the end + // of the day. + if self.cursor.prev == b':' { + self.discard_next = true; + } trace!("Candidate::Start\t"); - true + ParseAction::Consume } - _ => false, + _ => ParseAction::Skip, } } #[inline(always)] - fn parse_continue(&mut self, prev: u8, curr: u8, pos: usize) -> bool { - match curr { + fn parse_continue(&mut self) -> ParseAction<'a> { + match self.cursor.curr { // Enter arbitrary value mode - b'[' if prev == b'@' - || prev == b'-' - || prev == b' ' - || prev == b':' // Variant separator - || prev == b'/' // Modifier separator - || prev == b'!' - || prev == b'\0' => + b'[' if matches!( + self.cursor.prev, + b'@' | b'-' | b' ' | b':' | b'/' | b'!' | b'\0' + ) => { trace!("Arbitrary::Start\t"); self.in_arbitrary = true; - self.idx_arbitrary_start = pos; + self.idx_arbitrary_start = self.cursor.pos; } // Can't enter arbitrary value mode @@ -314,154 +442,305 @@ impl<'a> Extractor<'a> { b'[' => { trace!("Arbitrary::Skip_Start\t"); - return false; + return ParseAction::Skip; + } + + // A % can only appear at the end of the candidate itself. It can also only be after a + // digit 0-9. This covers the following cases: + // - from-15% + b'%' if self.cursor.prev.is_ascii_digit() => { + return match (self.cursor.at_end, self.cursor.next) { + // End of string == end of candidate == okay + (true, _) => ParseAction::Consume, + + // Looks like the end of a candidate == okay + (_, b' ' | b'\'' | b'"' | b'`') => ParseAction::Consume, + + // Otherwise, not a valid character in a candidate + _ => ParseAction::Skip, + }; + } + b'%' => return ParseAction::Skip, + + // < and > can only be part of a variant and only be the first or last character + b'<' | b'>' => { + // Can only be the first or last character + // E.g.: + // - :underline + // ^ + if self.cursor.pos == self.idx_start || self.cursor.pos == self.idx_last { + trace!("Candidate::Consume\t"); + } + // If it is in the middle, it can only be part of a stacked variant + // - dark::underline + // ^ + else if self.cursor.prev == b':' || self.cursor.next == b':' { + trace!("Candidate::Consume\t"); + } else { + return ParseAction::Skip; + } } // Allowed characters in the candidate itself // None of these can come after a closing bracket `]` - b'a'..=b'z' - | b'A'..=b'Z' - | b'0'..=b'9' - | b'-' - | b'_' - | b'(' - | b')' - | b'!' - | b'@' - | b'%' - if prev != b']' => + b'a'..=b'z' | b'A'..=b'Z' | b'0'..=b'9' | b'-' | b'_' | b'!' | b'@' + if self.cursor.prev != b']' => { /* TODO: The `b'@'` is necessary for custom separators like _@, maybe we can handle this in a better way... */ trace!("Candidate::Consume\t"); } + // A dot (.) can only appear in the candidate itself (not the arbitrary part), if the previous + // and next characters are both digits. This covers the following cases: + // - p-1.5 + b'.' if self.cursor.prev.is_ascii_digit() => match self.cursor.next { + next if next.is_ascii_digit() => { + trace!("Candidate::Consume\t"); + } + _ => return ParseAction::Skip, + }, + // Allowed characters in the candidate itself // These MUST NOT appear at the end of the candidate - b'/' | b':' | b'.' if pos + 1 < self.idx_last => { + b'/' | b':' if !self.cursor.at_end => { trace!("Candidate::Consume\t"); } - _ => return false, + _ => return ParseAction::Skip, } - true + ParseAction::Consume } #[inline(always)] - fn can_be_candidate(&mut self, c: u8) -> bool { + fn can_be_candidate(&mut self) -> bool { self.in_candidate && !self.in_arbitrary - && (0..=127).contains(&c) + && (0..=127).contains(&self.cursor.curr) && (self.idx_start == 0 || self.input[self.idx_start - 1] <= 127) } #[inline(always)] - fn handle_skip(&mut self, pos: usize) { + fn handle_skip(&mut self) { // In all other cases, we skip characters and reset everything so we can make new candidates trace!("Characters::Skip\t"); - self.idx_start = pos; - self.idx_end = pos; + self.idx_start = self.cursor.pos; + self.idx_end = self.cursor.pos; self.in_candidate = false; self.in_arbitrary = false; self.in_escape = false; } #[inline(always)] - fn parse_char(&mut self, prev: u8, curr: u8, pos: usize) -> bool { + fn parse_char(&mut self) -> ParseAction<'a> { if self.in_arbitrary { - self.parse_arbitrary(curr, pos) + self.parse_arbitrary() } else if self.in_candidate { - self.parse_continue(prev, curr, pos) - } else if self.parse_start(curr, pos) { + self.parse_continue() + } else if self.parse_start() == ParseAction::Consume { self.in_candidate = true; - self.idx_start = pos; - self.idx_end = pos; + self.idx_start = self.cursor.pos; + self.idx_end = self.cursor.pos; - true + ParseAction::Consume } else { - false + ParseAction::Skip } } #[inline(always)] - fn yield_candidate(&mut self, pos: usize, curr: u8, did_consume: bool) -> Option<&'a [u8]> { - // If we're still consuming characters, we keep going - // Only exception is if we've hit the end of the input - if did_consume && pos + 1 < self.idx_last { - return None; - } - - let candidate = if self.can_be_candidate(curr) { + fn yield_candidate(&mut self) -> ParseAction<'a> { + if self.can_be_candidate() { self.get_current_candidate() } else { - None - }; - - self.handle_skip(pos); - - candidate + ParseAction::Consume + } } #[inline(always)] - fn read(&mut self) -> (usize, u8) { - if self.pos == self.idx_last { - return (usize::MAX, 0); - } + fn restart(&mut self, pos: usize) { + trace!("Parser::Restart\t{}", pos); - let r = (self.pos, self.input[self.pos]); - self.pos += 1; + self.idx_start = pos; + self.idx_end = pos; + self.idx_arbitrary_start = 0; - r + self.in_arbitrary = false; + self.in_candidate = false; + self.in_escape = false; + + self.discard_next = false; + + self.quote_stack.clear(); + self.bracket_stack.clear(); + self.cursor.move_to(pos); } #[inline(always)] - fn parse_and_yield(&mut self) -> Option> { - let (pos, curr) = self.read(); - let did_consume = self.parse_char(self.prev, curr, pos); + fn without_surrounding(&self) -> Bracketing<'a> { + let range = self.idx_start..=self.idx_end; + let clipped = &self.input[range]; - if did_consume { - self.idx_end = pos; + Self::slice_surrounding(clipped) + .map(Bracketing::Included) + .or_else(|| { + if self.idx_start == 0 || self.idx_end + 1 == self.idx_last { + return None; + } + + let range = self.idx_start - 1..=self.idx_end + 1; + let clipped = &self.input[range]; + Self::slice_surrounding(clipped).map(Bracketing::Wrapped) + }) + .unwrap_or(Bracketing::None) + } + + #[inline(always)] + fn is_balanced(input: &[u8]) -> bool { + let mut depth = 0isize; + + for n in input { + match n { + b'[' | b'{' | b'(' => depth += 1, + b']' | b'}' | b')' => depth -= 1, + _ => continue, + } + + if depth < 0 { + return false; + } } - let candidate = self.yield_candidate(pos, curr, did_consume); - self.prev = curr; + depth == 0 + } - if curr == 0 { - None - } else { - Some(candidate) + #[inline(always)] + fn slice_surrounding(input: &[u8]) -> Option<&[u8]> { + let mut prev = None; + let mut input = input; + + loop { + let leading = input.first().unwrap_or(&0x00); + let trailing = input.last().unwrap_or(&0x00); + + let needed = matches!( + (leading, trailing), + (b'(', b')') + | (b'{', b'}') + | (b'[', b']') + | (b'"', b'"') + | (b'`', b'`') + | (b'\'', b'\'') + ); + + if needed { + prev = Some(input); + input = &input[1..input.len() - 1]; + continue; + } else if Self::is_balanced(input) && prev.is_some() { + return Some(input); + } else { + return prev; + } + } + } + + #[inline(always)] + fn parse_and_yield(&mut self) -> ParseAction<'a> { + trace!("Cursor {}", self.cursor); + + // Fast skipping of invalid characters + let can_skip_whitespace = false; // if self.opts.preserve_spaces_in_arbitrary { !self.in_arbitrary } else { true }; + if can_skip_whitespace { + if let Some(pos) = fast_skip(&self.cursor) { + trace!("FastSkip::Restart\t{}", pos); + return ParseAction::RestartAt(pos); + } + } + + let action = self.parse_char(); + + match action { + ParseAction::RestartAt(_) => return action, + ParseAction::Consume => { + self.idx_end = self.cursor.pos; + + // If we're still consuming characters, we keep going + // Only exception is if we've hit the end of the input + if !self.cursor.at_end { + return action; + } + } + _ => {} + } + + let action = self.yield_candidate(); + + match (&action, self.cursor.curr) { + (ParseAction::RestartAt(_), _) => action, + (_, 0x00) => ParseAction::Done, + (ParseAction::SingleCandidate(candidate), _) => self.generate_slices(candidate), + _ => ParseAction::RestartAt(self.cursor.pos + 1), + } + } + + /// Peek inside `[]`, `{}`, and `()` pairs + /// to look for an additional candidate + #[inline(always)] + fn generate_slices(&mut self, candidate: &'a [u8]) -> ParseAction<'a> { + match self.without_surrounding() { + Bracketing::None => ParseAction::SingleCandidate(candidate), + Bracketing::Included(slicable) if slicable == candidate => { + ParseAction::SingleCandidate(candidate) + } + Bracketing::Included(slicable) | Bracketing::Wrapped(slicable) => { + let parts = vec![candidate, slicable]; + let parts = parts + .into_iter() + .filter(|v| !v.is_empty()) + .collect::>(); + + ParseAction::MultipleCandidates(parts) + } } } } impl<'a> Iterator for Extractor<'a> { - type Item = &'a [u8]; + type Item = Vec<&'a [u8]>; fn next(&mut self) -> Option { - if self.pos == self.idx_last { + if self.cursor.at_end { return None; } loop { - let candidate = self.parse_and_yield()?; - if candidate.is_some() { - return candidate; + let result = self.parse_and_yield(); + + // Cursor control + match result { + ParseAction::RestartAt(pos) => self.restart(pos), + _ => self.cursor.advance_by(1), } - let candidate = self.parse_and_yield()?; - if candidate.is_some() { - return candidate; + // Candidate state control + match result { + ParseAction::SingleCandidate(_) => self.handle_skip(), + ParseAction::MultipleCandidates(_) => self.handle_skip(), + _ => {} } - let candidate = self.parse_and_yield()?; - if candidate.is_some() { - return candidate; - } - - let candidate = self.parse_and_yield()?; - if candidate.is_some() { - return candidate; - } + // Iterator results + return match result { + ParseAction::SingleCandidate(candidate) => Some(vec![candidate]), + ParseAction::MultipleCandidates(candidates) => Some(candidates), + ParseAction::Done => None, + _ => continue, + }; } } } @@ -542,8 +821,7 @@ mod test { fn it_should_not_keep_spaces() { let candidates = run("bg-[rgba(0, 0, 0)]", false); - // TODO: This should be empty, because the arbitrary value is unbalanced. - assert!(candidates.is_empty()); + assert_eq!(candidates, vec!["rgba"]); } #[test] @@ -564,7 +842,7 @@ mod test { assert!(candidates.is_empty()); let candidates = run("[something]", false); - assert!(candidates.is_empty()); + assert_eq!(candidates, vec!["something"]); let candidates = run("[color:red]/dark", false); assert!(candidates.is_empty()); @@ -573,7 +851,7 @@ mod test { assert!(candidates.is_empty()); let candidates = run(" [feature(slice_as_chunks)]", false); - assert!(candidates.is_empty()); + assert_eq!(candidates, vec!["feature(slice_as_chunks)"]); let candidates = run("![feature(slice_as_chunks)]", false); assert!(candidates.is_empty()); @@ -632,6 +910,7 @@ mod test { assert_eq!( candidates, vec![ + "div", "class", r#"dark:lg:hover:[&>*]:underline"#, r#"[&_.foo\_\_bar]:hover:underline"#, @@ -646,7 +925,7 @@ mod test { assert_eq!( candidates, vec![ - // "p", // TODO: This is missing. Maybe? + "p", "class", "text-sm", "text-blue-700", @@ -655,22 +934,22 @@ mod test { "software", "update", "is", - "available.", + "available", "See", // "what", // what is dropped because it is followed by the fancy: ’ // "s", // s is dropped because it is preceeded by the fancy: ’ // "new", // Already seen "in", "version", - "p", // Hmm, becuse "

" ] ); } #[test] fn ignores_arbitrary_property_ish_things() { + // FIXME: () are only valid in an arbitrary let candidates = run(" [feature(slice_as_chunks)]", false); - assert!(candidates.is_empty()); + assert_eq!(candidates, vec!["feature(slice_as_chunks)",]); } #[test] @@ -697,7 +976,7 @@ mod test { #[test] fn bad_001() { let candidates = run("[杛杛]/", false); - assert!(candidates.is_empty()) + assert_eq!(candidates, vec!["杛杛"]); } #[test] @@ -708,7 +987,117 @@ mod test { #[test] fn bad_003() { + // TODO: This seems… wrong let candidates = run(r"[𕤵:]", false); - assert!(candidates.is_empty()); + assert_eq!(candidates, vec!["𕤵", "𕤵:"]); + } + + #[test] + fn classes_in_js_arrays() { + let candidates = run( + r#"let classes = ['bg-black', 'hover:px-0.5', 'text-[13px]', '[--my-var:1_/_2]', '[.foo_&]:px-[0]', '[.foo_&]:[color:red]']">"#, + false, + ); + assert_eq!( + candidates, + vec![ + "let", + "classes", + "bg-black", + "hover:px-0.5", + "text-[13px]", + "[--my-var:1_/_2]", + "--my-var:1_/_2", + "[.foo_&]:px-[0]", + "[.foo_&]:[color:red]", + ] + ); + } + + #[test] + fn classes_as_object_keys() { + let candidates = run( + r#"
"#, + false, + ); + assert_eq!( + candidates, + vec!["div", "underline", "isActive", "px-1.5", "isOnline"] + ); + } + + #[test] + fn multiple_nested_candidates() { + let candidates = run(r#"{color:red}"#, false); + assert_eq!(candidates, vec!["color:red"]); + } + + #[test] + fn percent_ended_candidates() { + let candidates = run( + r#""#, + false, + ); + assert_eq!( + candidates, + vec![ + "!--", + "This", + "should", + "work", + "underline", + "from-50%", + "flex", + "--" + ] + ); + } + + #[test] + fn ruby_percent_formatted_strings() { + let candidates = run(r#"%w[hover:flex]"#, false); + assert_eq!(candidates, vec!["w", "hover:flex"]); + } + + #[test] + fn candidate_slicing() { + let result = Extractor::slice_surrounding(&b".foo_&]:px-[0"[..]) + .map(std::str::from_utf8) + .transpose() + .unwrap(); + assert_eq!(result, None); + + let result = Extractor::slice_surrounding(&b"[.foo_&]:px-[0]"[..]) + .map(std::str::from_utf8) + .transpose() + .unwrap(); + assert_eq!(result, Some("[.foo_&]:px-[0]")); + + let result = Extractor::slice_surrounding(&b"{[.foo_&]:px-[0]}"[..]) + .map(std::str::from_utf8) + .transpose() + .unwrap(); + assert_eq!(result, Some("[.foo_&]:px-[0]")); + + let result = Extractor::slice_surrounding(&b"![foo:bar]"[..]) + .map(std::str::from_utf8) + .transpose() + .unwrap(); + assert_eq!(result, None); + + let result = Extractor::slice_surrounding(&b"[\"pt-1.5\"]"[..]) + .map(std::str::from_utf8) + .transpose() + .unwrap(); + assert_eq!(result, Some("pt-1.5")); + + let count = 1_000; + let crazy = format!("{}[.foo_&]:px-[0]{}", "[".repeat(count), "]".repeat(count)); + + let result = Extractor::slice_surrounding(crazy.as_bytes()) + .map(std::str::from_utf8) + .transpose() + .unwrap(); + assert_eq!(result, Some("[.foo_&]:px-[0]")); } } diff --git a/src/featureFlags.js b/src/featureFlags.js index b4eca6665..632049605 100644 --- a/src/featureFlags.js +++ b/src/featureFlags.js @@ -9,7 +9,7 @@ let defaults = { logicalSiblingUtilities: false, } -let featureFlags = { +export let featureFlags = { future: [ 'hoverOnlyWhenSupported', 'respectDefaultRingColorOpacity', diff --git a/src/util/normalizeConfig.js b/src/util/normalizeConfig.js index 3cb98511c..edab3f466 100644 --- a/src/util/normalizeConfig.js +++ b/src/util/normalizeConfig.js @@ -1,4 +1,4 @@ -import { flagEnabled } from '../featureFlags' +import { flagEnabled, featureFlags } from '../featureFlags' import log, { dim } from './log' export function normalizeConfig(config) { @@ -294,6 +294,21 @@ export function normalizeConfig(config) { })(), } + // Force disable the `oxideParser` feature flag when using unsupported features. + // TODO: Remove once we have prefix or separator support in the oxide parser. + if (config.prefix !== '' || config.separator !== ':') { + if (config.experimental === 'all') { + config.experimental = {} + for (let key in featureFlags.experimental) { + config.experimental[key] = true + } + } else { + config.experimental = config.experimental ?? {} + } + + config.experimental.oxideParser = false + } + // Validate globs to prevent bogus globs. // E.g.: `./src/*.{html}` is invalid, the `{html}` should just be `html` if (config.content.files !== 'auto') { diff --git a/tests/animations.test.js b/tests/animations.test.js index ab2467476..dd05c6eba 100644 --- a/tests/animations.test.js +++ b/tests/animations.test.js @@ -187,7 +187,7 @@ test('with dots in the name', () => { content: [ { raw: html` -
+
`, }, @@ -195,11 +195,11 @@ test('with dots in the name', () => { theme: { extend: { keyframes: { - 'zoom-.5': { to: { transform: 'scale(0.5)' } }, + 'zoom-0.5': { to: { transform: 'scale(0.5)' } }, 'zoom-1.5': { to: { transform: 'scale(1.5)' } }, }, animation: { - 'zoom-.5': 'zoom-.5 2s', + 'zoom-0.5': 'zoom-0.5 2s', 'zoom-1.5': 'zoom-1.5 2s', }, }, @@ -208,19 +208,22 @@ test('with dots in the name', () => { return run('@tailwind utilities', config).then((result) => { expect(result.css).toMatchFormattedCss(css` - @keyframes zoom-\.5 { + @keyframes zoom-0\.5 { to { transform: scale(0.5); } } - .animate-zoom-\.5 { - animation: 2s zoom-\.5; + + .animate-zoom-0\.5 { + animation: 2s zoom-0\.5; } + @keyframes zoom-1\.5 { to { transform: scale(1.5); } } + .animate-zoom-1\.5 { animation: 2s zoom-1\.5; } diff --git a/tests/arbitrary-variants.test.js b/tests/arbitrary-variants.test.js index da69272f1..3bb8133c8 100644 --- a/tests/arbitrary-variants.test.js +++ b/tests/arbitrary-variants.test.js @@ -1,4 +1,5 @@ import { run, html, css, defaults } from './util/run' +import { flagEnabled } from '../src/featureFlags' test('basic arbitrary variants', () => { let config = { @@ -612,24 +613,44 @@ it('should support aria variants', () => { ` return run(input, config).then((result) => { - expect(result.css).toMatchFormattedCss(css` - .underline, - .aria-checked\:underline[aria-checked='true'], - .aria-\[labelledby\=\'a_b\'\]\:underline[aria-labelledby='a b'], - .aria-\[sort\=ascending\]\:underline[aria-sort='ascending'], - .group\/foo[aria-checked='true'] .group-aria-checked\/foo\:underline, - .group[aria-checked='true'] .group-aria-checked\:underline, - .group[aria-labelledby='a b'] .group-aria-\[labelledby\=\'a_b\'\]\:underline, - .group\/foo[aria-sort='ascending'] .group-aria-\[sort\=ascending\]\/foo\:underline, - .group[aria-sort='ascending'] .group-aria-\[sort\=ascending\]\:underline, - .peer\/foo[aria-checked='true'] ~ .peer-aria-checked\/foo\:underline, - .peer[aria-checked='true'] ~ .peer-aria-checked\:underline, - .peer[aria-labelledby='a b'] ~ .peer-aria-\[labelledby\=\'a_b\'\]\:underline, - .peer\/foo[aria-sort='ascending'] ~ .peer-aria-\[sort\=ascending\]\/foo\:underline, - .peer[aria-sort='ascending'] ~ .peer-aria-\[sort\=ascending\]\:underline { - text-decoration-line: underline; - } - `) + expect(result.css).toMatchFormattedCss( + flagEnabled(config, 'oxideParser') + ? css` + .aria-checked\:underline[aria-checked='true'], + .aria-\[labelledby\=\'a_b\'\]\:underline[aria-labelledby='a b'], + .aria-\[sort\=ascending\]\:underline[aria-sort='ascending'], + .group\/foo[aria-checked='true'] .group-aria-checked\/foo\:underline, + .group[aria-checked='true'] .group-aria-checked\:underline, + .group[aria-labelledby='a b'] .group-aria-\[labelledby\=\'a_b\'\]\:underline, + .group\/foo[aria-sort='ascending'] .group-aria-\[sort\=ascending\]\/foo\:underline, + .group[aria-sort='ascending'] .group-aria-\[sort\=ascending\]\:underline, + .peer\/foo[aria-checked='true'] ~ .peer-aria-checked\/foo\:underline, + .peer[aria-checked='true'] ~ .peer-aria-checked\:underline, + .peer[aria-labelledby='a b'] ~ .peer-aria-\[labelledby\=\'a_b\'\]\:underline, + .peer\/foo[aria-sort='ascending'] ~ .peer-aria-\[sort\=ascending\]\/foo\:underline, + .peer[aria-sort='ascending'] ~ .peer-aria-\[sort\=ascending\]\:underline { + text-decoration-line: underline; + } + ` + : css` + .underline, + .aria-checked\:underline[aria-checked='true'], + .aria-\[labelledby\=\'a_b\'\]\:underline[aria-labelledby='a b'], + .aria-\[sort\=ascending\]\:underline[aria-sort='ascending'], + .group\/foo[aria-checked='true'] .group-aria-checked\/foo\:underline, + .group[aria-checked='true'] .group-aria-checked\:underline, + .group[aria-labelledby='a b'] .group-aria-\[labelledby\=\'a_b\'\]\:underline, + .group\/foo[aria-sort='ascending'] .group-aria-\[sort\=ascending\]\/foo\:underline, + .group[aria-sort='ascending'] .group-aria-\[sort\=ascending\]\:underline, + .peer\/foo[aria-checked='true'] ~ .peer-aria-checked\/foo\:underline, + .peer[aria-checked='true'] ~ .peer-aria-checked\:underline, + .peer[aria-labelledby='a b'] ~ .peer-aria-\[labelledby\=\'a_b\'\]\:underline, + .peer\/foo[aria-sort='ascending'] ~ .peer-aria-\[sort\=ascending\]\/foo\:underline, + .peer[aria-sort='ascending'] ~ .peer-aria-\[sort\=ascending\]\:underline { + text-decoration-line: underline; + } + ` + ) }) }) @@ -669,24 +690,44 @@ it('should support data variants', () => { ` return run(input, config).then((result) => { - expect(result.css).toMatchFormattedCss(css` - .underline, - .data-checked\:underline[data-ui~='checked'], - .data-\[foo\=\'bar_baz\'\]\:underline[data-foo='bar baz'], - .data-\[position\=top\]\:underline[data-position='top'], - .group\/foo[data-ui~='checked'] .group-data-checked\/foo\:underline, - .group[data-ui~='checked'] .group-data-checked\:underline, - .group[data-foo='bar baz'] .group-data-\[foo\=\'bar_baz\'\]\:underline, - .group\/foo[data-position='top'] .group-data-\[position\=top\]\/foo\:underline, - .group[data-position='top'] .group-data-\[position\=top\]\:underline, - .peer\/foo[data-ui~='checked'] ~ .peer-data-checked\/foo\:underline, - .peer[data-ui~='checked'] ~ .peer-data-checked\:underline, - .peer[data-foo='bar baz'] ~ .peer-data-\[foo\=\'bar_baz\'\]\:underline, - .peer\/foo[data-position='top'] ~ .peer-data-\[position\=top\]\/foo\:underline, - .peer[data-position='top'] ~ .peer-data-\[position\=top\]\:underline { - text-decoration-line: underline; - } - `) + expect(result.css).toMatchFormattedCss( + flagEnabled(config, 'oxideParser') + ? css` + .data-checked\:underline[data-ui~='checked'], + .data-\[foo\=\'bar_baz\'\]\:underline[data-foo='bar baz'], + .data-\[position\=top\]\:underline[data-position='top'], + .group\/foo[data-ui~='checked'] .group-data-checked\/foo\:underline, + .group[data-ui~='checked'] .group-data-checked\:underline, + .group[data-foo='bar baz'] .group-data-\[foo\=\'bar_baz\'\]\:underline, + .group\/foo[data-position='top'] .group-data-\[position\=top\]\/foo\:underline, + .group[data-position='top'] .group-data-\[position\=top\]\:underline, + .peer\/foo[data-ui~='checked'] ~ .peer-data-checked\/foo\:underline, + .peer[data-ui~='checked'] ~ .peer-data-checked\:underline, + .peer[data-foo='bar baz'] ~ .peer-data-\[foo\=\'bar_baz\'\]\:underline, + .peer\/foo[data-position='top'] ~ .peer-data-\[position\=top\]\/foo\:underline, + .peer[data-position='top'] ~ .peer-data-\[position\=top\]\:underline { + text-decoration-line: underline; + } + ` + : css` + .underline, + .data-checked\:underline[data-ui~='checked'], + .data-\[foo\=\'bar_baz\'\]\:underline[data-foo='bar baz'], + .data-\[position\=top\]\:underline[data-position='top'], + .group\/foo[data-ui~='checked'] .group-data-checked\/foo\:underline, + .group[data-ui~='checked'] .group-data-checked\:underline, + .group[data-foo='bar baz'] .group-data-\[foo\=\'bar_baz\'\]\:underline, + .group\/foo[data-position='top'] .group-data-\[position\=top\]\/foo\:underline, + .group[data-position='top'] .group-data-\[position\=top\]\:underline, + .peer\/foo[data-ui~='checked'] ~ .peer-data-checked\/foo\:underline, + .peer[data-ui~='checked'] ~ .peer-data-checked\:underline, + .peer[data-foo='bar baz'] ~ .peer-data-\[foo\=\'bar_baz\'\]\:underline, + .peer\/foo[data-position='top'] ~ .peer-data-\[position\=top\]\/foo\:underline, + .peer[data-position='top'] ~ .peer-data-\[position\=top\]\:underline { + text-decoration-line: underline; + } + ` + ) }) }) diff --git a/tests/basic-usage.test.js b/tests/basic-usage.test.js index 9115cacd0..d32ac43c9 100644 --- a/tests/basic-usage.test.js +++ b/tests/basic-usage.test.js @@ -1,6 +1,7 @@ import fs from 'fs' import path from 'path' import { run, html, css, defaults } from './util/run' +import { flagEnabled } from '../src/featureFlags' test('basic usage', () => { let config = { @@ -877,11 +878,15 @@ test('should not crash when group names contain special characters', () => { ` return run(input, config).then((result) => { - expect(result.css).toMatchFormattedCss(css` - .group\/\$\{id\}:hover .group-hover\/\$\{id\}\:visible { - visibility: visible; - } - `) + if (flagEnabled(config, 'oxideParser')) { + expect(result.css).toMatchFormattedCss(css``) + } else { + expect(result.css).toMatchFormattedCss(css` + .group\/\$\{id\}:hover .group-hover\/\$\{id\}\:visible { + visibility: visible; + } + `) + } }) }) @@ -927,12 +932,20 @@ test('detects quoted arbitrary values containing a slash', async () => { let result = await run(input, config) - expect(result.css).toMatchFormattedCss(css` - .hidden, - .group[href^='/'] .group-\[\[href\^\=\'\/\'\]\]\:hidden { - display: none; - } - `) + expect(result.css).toMatchFormattedCss( + flagEnabled(config, 'oxideParser') + ? css` + .group[href^='/'] .group-\[\[href\^\=\'\/\'\]\]\:hidden { + display: none; + } + ` + : css` + .hidden, + .group[href^='/'] .group-\[\[href\^\=\'\/\'\]\]\:hidden { + display: none; + } + ` + ) }) test('handled quoted arbitrary values containing escaped spaces', async () => { @@ -950,10 +963,18 @@ test('handled quoted arbitrary values containing escaped spaces', async () => { let result = await run(input, config) - expect(result.css).toMatchFormattedCss(css` - .hidden, - .group[href^=' bar'] .group-\[\[href\^\=\'_bar\'\]\]\:hidden { - display: none; - } - `) + expect(result.css).toMatchFormattedCss( + flagEnabled(config, 'oxideParser') + ? css` + .group[href^=' bar'] .group-\[\[href\^\=\'_bar\'\]\]\:hidden { + display: none; + } + ` + : css` + .hidden, + .group[href^=' bar'] .group-\[\[href\^\=\'_bar\'\]\]\:hidden { + display: none; + } + ` + ) }) diff --git a/tests/parse-candidate-strings.test.js b/tests/parse-candidate-strings.test.js new file mode 100644 index 000000000..63d4f82ab --- /dev/null +++ b/tests/parse-candidate-strings.test.js @@ -0,0 +1,423 @@ +import { parseCandidateStrings, IO, Parsing } from '@tailwindcss/oxide' +import { defaultExtractor as createDefaultExtractor } from '../src/lib/defaultExtractor' + +let html = String.raw + +let defaultExtractor = createDefaultExtractor({ tailwindConfig: { separator: ':' } }) + +function regexParser(str) { + return defaultExtractor(str) +} + +function oxideParser(str) { + return parseCandidateStrings( + [{ content: str, extension: 'html' }], + IO.Sequential | Parsing.Parallel + ) +} + +function templateTable(classes) { + classes = classes.concat( + // Variants + classes.flatMap((c) => { + return [ + // Simple variant + `hover:${c}`, + + // Combined variant + `focus:hover:${c}`, + + // Variant with dashes + `group-hover:group-focus:${c}`, + + // With special characters + `:${c}`, + + // With arbitrary values + `min-[300px]:${c}`, + + // With arbitrary values + `[@media(hover:hover)]:${c}`, + + // With parent selector + `[.foo_&]:${c}`, + ] + }) + ) + + let classString = classes.join(' ') + let singleQuoteArraySyntax = `'${classes.join("', '")}'` + + return [ + ['Plain', classString], + ['HTML', html`
`], + + ['JavaScript variable', `let foo = "${classString}"`], + ['JavaScript expression', `document.body.classList.add(['${classString}'].join(' '))`], + ['JavaScript object key', `let foo = {'${classString}': true}`], + + ['JSX basic', html`
`], + ['JSX with JavaScript expression', html`
`], + + ['Vue basic', html`
`], + ['Vue array (single quote)', html`
`], + ['Vue object (single quote)', html`
`], + + ['Markdown code fences', ``], + ].map(([name, template]) => [name, template, classes]) +} + +describe.each([ + ['Regex', regexParser], + ['Oxide', oxideParser], +])('%s parser', (_, parse) => { + describe('basic utility classes', () => { + let classes = [ + // One word classes + 'underline', + + // With dashes + 'text-center', + 'pointer-events-none', + + // With numbers + 'px-4', + + // With special characters + 'px-1.5', + 'translate-x-1/2', + 'from-50%', + + // With negative signs + '-translate-x-full', + + // With halves and negative signs + '-translate-x-1/2', + ] + + test.each(templateTable(classes))('%# — %s', (_, template, classes) => { + let extractions = parse(template) + + for (let c of classes) { + expect(extractions).toContain(c) + } + }) + }) + + describe('utility classes with arbitrary values', () => { + let classes = [ + // With simple number + 'px-[0]', + 'px-[0.5]', + + // With number and unit + 'px-[123px]', + 'px-[123.45px]', + + // With special symbols + 'px-[#bada55]', + // ^ + 'px-[color:#bada55]', + // ^^ + 'content-[>]', + // ^ + 'content-[<]', + // ^ + + // With functions and math expressions + 'px-[calc(100%-1rem)]', + 'px-[theme(spacing.1)]', + 'px-[theme(spacing[1.5])]', + + // With spaces (replaced by `_`) + 'bg-[rgb(255_0_0)]', + + // Examples with combinations + 'w-[calc(100%_-_theme("spacing[1.5]))"]', + 'fill-[oklab(59.69%_0.1007_0.1191_/_0.5)]/[33.7%]', + 'fill-[color:oklab(59.69%_0.1007_0.1191_/_0.5)]/[33.7%]', + 'shadow-[inset_0_-3em_3em_rgba(0,_0,_0,_0.1),_0_0_0_2px_rgb(255,_255,_255),_0.3em_0.3em_1em_rgba(0,_0,_0,_0.3)]', + ] + + test.each(templateTable(classes))('%# — %s', (_, template, classes) => { + let extractions = parse(template) + + for (let c of classes) { + expect(extractions).toContain(c) + } + }) + }) + + describe('utility classes with modifiers', () => { + let classes = [ + // With simple modifiers + 'bg-red-500/50', + + // With arbitrary modifiers + 'bg-red-500/[0.5]', + 'bg-red-500/[50%]', + 'bg-red-500/[var(--opacity)]', + + // With spces (replaced by `_`) + 'bg-red-500/[var(--opacity,_50%)]', + ] + + test.each(templateTable(classes))('%# — %s', (_, template, classes) => { + let extractions = parse(template) + + for (let c of classes) { + expect(extractions).toContain(c) + } + }) + }) + + describe('utility classes with important modifier', () => { + let classes = ['!bg-red-500', '!bg-[#bada55]', '![display:flex]', '!-translate-x-1/2'] + + test.each(templateTable(classes))('%# — %s', (_, template, classes) => { + let extractions = parse(template) + + for (let c of classes) { + expect(extractions).toContain(c) + } + }) + }) + + describe('arbitrary properties', () => { + let classes = [ + // With simple arbitrary property + '[display:flex]', + ] + + test.each(templateTable(classes))('%# — %s', (_, template, classes) => { + let extractions = parse(template) + + for (let c of classes) { + expect(extractions).toContain(c) + } + }) + }) + + describe('arbitrary values with quotes', () => { + test('arbitrary values with single quotes', async () => { + let extractions = parse(` +
+ `) + + expect(extractions).toContain(`content-['hello_world']`) + expect(extractions).toContain(`hover:content-['hello_world_2']`) + expect(extractions).toContain(`hover:focus:content-['hello_world_3']`) + }) + + test('arbitrary values with double quotes', async () => { + let extractions = parse(` +
+ `) + + expect(extractions).toContain(`content-["hello_world"]`) + expect(extractions).toContain(`hover:content-["hello_world_2"]`) + expect(extractions).toContain(`hover:focus:content-["hello_world_3"]`) + }) + + test('arbitrary values with some single quoted values', async () => { + let extractions = parse(` +
+ `) + + expect(extractions).toContain(`font-['Open_Sans',_system-ui,_sans-serif]`) + expect(extractions).toContain(`hover:font-['Proxima_Nova',_system-ui,_sans-serif]`) + expect(extractions).toContain(`hover:focus:font-['Inter_var',_system-ui,_sans-serif]`) + }) + + test('arbitrary values with some double quoted values', async () => { + let extractions = parse(` +
+ `) + + expect(extractions).toContain(`font-["Open_Sans",_system-ui,_sans-serif]`) + expect(extractions).toContain(`hover:font-["Proxima_Nova",_system-ui,_sans-serif]`) + expect(extractions).toContain(`hover:focus:font-["Inter_var",_system-ui,_sans-serif]`) + }) + + test('arbitrary values with escaped underscores', async () => { + let extractions = parse(` +
+ `) + + expect(extractions).toContain(`content-['hello\\_world']`) + expect(extractions).toContain(`hover:content-['hello\\_world\\_2']`) + expect(extractions).toContain(`hover:focus:content-['hello\\_world\\_3']`) + }) + + test('arbitrary values with angle brackets in single quotes', async () => { + let extractions = parse(` +
+ `) + + expect(extractions).toContain(`content-['>']`) + expect(extractions).toContain(`hover:content-['<']`) + expect(extractions).toContain(`hover:focus:content-['>']`) + }) + + test('arbitrary values with angle brackets in double quotes', async () => { + let extractions = parse(` +
"] hover:content-["<"] hover:focus:content-[">"]">
+ `) + + expect(extractions).toContain(`content-[">"]`) + expect(extractions).toContain(`hover:content-["<"]`) + expect(extractions).toContain(`hover:focus:content-[">"]`) + }) + + test('arbitrary values with theme lookup using quotes', () => { + let extractions = parse(` +

+ `) + + expect(extractions).toContain(`[--y:theme('colors.blue.500')]`) + expect(extractions).toContain(`[color:var(--y)]`) + }) + }) + + test('special characters', async () => { + let extractions = parse(` +
+ `) + + expect(extractions).toContain(`:font-bold`) + }) + + test('with single quotes array within template literal', async () => { + let extractions = parse(`
`) + + expect(extractions).toContain('pr-1.5') + }) + + test('with double quotes array within template literal', async () => { + let extractions = parse(`
`) + + expect(extractions).toContain('pr-1.5') + }) + + test('with single quotes array within function', async () => { + let extractions = parse(`document.body.classList.add(['pl-1.5'].join(" "));`) + + expect(extractions).toContain('pl-1.5') + }) + + test('with double quotes array within function', async () => { + let extractions = parse(`document.body.classList.add(["pl-1.5"].join(" "));`) + + expect(extractions).toContain('pl-1.5') + }) + + test('with angle brackets', async () => { + let extractions = parse( + `
test
` + ) + + expect(extractions).toContain('bg-blue-200') + expect(extractions).toContain('shadow-xl') + expect(extractions).not.toContain('>shadow-xl') + expect(extractions).not.toContain('shadow-xl<') + }) + + test('markdown code fences', async () => { + let extractions = parse('') + + expect(extractions).toContain('font-bold') + expect(extractions).toContain('font-normal') + expect(extractions).not.toContain('.font-bold') + expect(extractions).not.toContain('.font-normal') + }) + + test('classes in slim templates', async () => { + let extractions = parse(` + p.bg-red-500.text-sm + 'This is a paragraph + small.italic.text-gray-500 + '(Look mom, no closing tag!) + `) + + expect(extractions).toContain('bg-red-500') + expect(extractions).toContain('text-sm') + expect(extractions).toContain('italic') + expect(extractions).toContain('text-gray-500') + }) + + test('multi-word + arbitrary values + quotes', async () => { + let extractions = parse(` + grid-cols-['repeat(2)'] + `) + + expect(extractions).toContain(`grid-cols-['repeat(2)']`) + }) + + test('a lot of data', () => { + let extractions = parse('underline '.repeat(2 ** 17)) + + expect(extractions).toContain(`underline`) + }) + + test('ruby percent string array', () => { + let extractions = parse('%w[text-[#bada55]]') + + expect(extractions).toContain(`text-[#bada55]`) + }) + + test('arbitrary properties followed by square bracketed stuff', () => { + let extractions = parse( + '
[foo]
' + ) + + expect(extractions).toContain(`[display:inherit]`) + }) + + describe('Vue', () => { + test('Class object syntax', () => { + let extractions = parse( + `[foo]` + ) + + expect(extractions).toContain(`underline`, `font-bold`) + }) + + test('Class array syntax', () => { + // With leading space + let extractions = parse( + `
[foo]
` + ) + + expect(extractions).toContain(`underline`, `font-bold`, `flex`, `block`) + + // Without leading space + extractions = parse( + `
[foo]
` + ) + + expect(extractions).toContain(`underline`, `font-bold`, `flex`, `block`) + }) + }) + + describe('anti-test', () => { + it('should not parse candidates without spaces in object-like syntax', () => { + let extractions = parse(html`
`) + + // The oxide parser _does_ allow this! + if (parse !== oxideParser) { + expect(extractions).not.toContain(`underline:isActive`) + expect(extractions).not.toContain(`flex:isOnline`) + expect(extractions).not.toContain(`md:bold`) + } + + expect(extractions).not.toContain(`underline`) + expect(extractions).not.toContain(`isActive`) + expect(extractions).not.toContain(`flex`) + expect(extractions).not.toContain(`isOnline`) + expect(extractions).not.toContain(`md`) + expect(extractions).not.toContain(`bold`) + }) + }) +})