From 64202758ec8693dd18951c554728bb5c6b7e5eeb Mon Sep 17 00:00:00 2001 From: sigoden Date: Tue, 31 Oct 2023 18:34:38 +0800 Subject: chore: move split_line_* from repl/cmd.rs to utils/ (#190) --- src/utils/mod.rs | 2 + src/utils/split_line.rs | 216 ++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 218 insertions(+) create mode 100644 src/utils/split_line.rs (limited to 'src/utils') diff --git a/src/utils/mod.rs b/src/utils/mod.rs index db53fd6..9e999d4 100644 --- a/src/utils/mod.rs +++ b/src/utils/mod.rs @@ -1,5 +1,7 @@ +mod split_line; mod tiktoken; +pub use self::split_line::*; pub use self::tiktoken::cl100k_base_singleton; use chrono::prelude::*; diff --git a/src/utils/split_line.rs b/src/utils/split_line.rs new file mode 100644 index 0000000..105c5f3 --- /dev/null +++ b/src/utils/split_line.rs @@ -0,0 +1,216 @@ +pub fn split_line_sematic(text: &str) -> Option<(String, String)> { + let mut balance: Vec = Vec::new(); + let chars: Vec = text.chars().collect(); + let mut index = 0; + let len = chars.len(); + while index < len - 1 { + let ch = chars[index]; + if balance.is_empty() + && ((matches!(ch, ',' | '.' | ';') && chars[index + 1].is_whitespace()) + || matches!(ch, '，' | '。' | '；')) + { + let (output, remain) = chars.split_at(index + 1); + return Some((output.iter().collect(), remain.iter().collect())); + } + if index + 2 < len && do_balance(&mut balance, &chars[index..=index + 2]) { + index += 3; + continue; + } + if do_balance(&mut balance, &chars[index..=index + 1]) { + index += 2; + continue; + } + do_balance(&mut balance, &chars[index..=index]); + index += 1; + } + + None +} + +pub fn split_line_tail(text: &str) -> (&str, &str) { + if let Some((head, tail)) = text.rsplit_once('\n') { + (head, tail) + } else { + ("", text) + } +} + +pub fn spaces(n: usize) -> String { + " ".repeat(n) +} + +#[derive(Debug, Clone, Copy, Eq, PartialEq)] +enum Kind { + ParentheseStart, + ParentheseEnd, + BracketStart, + BracketEnd, + Asterisk, + Asterisk2, + SingleQuota, + DoubleQuota, + Tilde, + Tilde2, + Backtick, + Backtick3, +} + +impl Kind { + fn from_chars(chars: &[char]) -> Option { + let kind = match chars.len() { + 1 => match chars[0] { + '(' => Self::ParentheseStart, + ')' => Self::ParentheseEnd, + '[' => Self::BracketStart, + ']' => Self::BracketEnd, + '*' => Self::Asterisk, + '\'' => Self::SingleQuota, + '"' => Self::DoubleQuota, + '~' => Self::Tilde, + '`' => Self::Backtick, + _ => return None, + }, + 2 if chars[0] == chars[1] => match chars[0] { + '*' => Self::Asterisk2, + '~' => Self::Tilde2, + _ => return None, + }, + 3 => { + if chars == ['`', '`', '`'] { + Self::Backtick3 + } else { + return None; + } + } + _ => return None, + }; + Some(kind) + } +} + +fn do_balance(balance: &mut Vec, chars: &[char]) -> bool { + Kind::from_chars(chars).map_or(false, |kind| { + let last = balance.last(); + match (kind, last) { + (Kind::ParentheseEnd, Some(&Kind::ParentheseStart)) + | (Kind::BracketEnd, Some(&Kind::BracketStart)) + | (Kind::Asterisk, Some(&Kind::Asterisk)) + | (Kind::Asterisk2, Some(&Kind::Asterisk2)) + | (Kind::SingleQuota, Some(&Kind::SingleQuota)) + | (Kind::DoubleQuota, Some(&Kind::DoubleQuota)) + | (Kind::Tilde, Some(&Kind::Tilde)) + | (Kind::Tilde2, Some(&Kind::Tilde2)) + | (Kind::Backtick, Some(&Kind::Backtick)) + | (Kind::Backtick3, Some(&Kind::Backtick3)) => { + balance.pop(); + true + } + ( + Kind::ParentheseStart + | Kind::BracketStart + | Kind::Asterisk + | Kind::Asterisk2 + | Kind::SingleQuota + | Kind::DoubleQuota + | Kind::Tilde + | Kind::Tilde2 + | Kind::Backtick + | Kind::Backtick3, + _, + ) => { + balance.push(kind); + true + } + _ => false, + } + }) +} + +#[cfg(test)] +mod tests { + use super::*; + + macro_rules! assert_split_line { + ($a:literal, $b:literal, true) => { + assert_eq!( + split_line_sematic(&format!("{}{}", $a, $b)), + Some(($a.into(), $b.into())) + ); + }; + ($a:literal, $b:literal, false) => { + assert_eq!(split_line_sematic(&format!("{}{}", $a, $b)), None); + }; + } + + #[test] + fn test_split_line() { + assert_split_line!( + "Wikipedia is a free online encyclopedia,", + " that anyone can edit,", + true + ); + assert_split_line!( + "Wikipedia is a free online encyclopedia.", + " that anyone can edit,", + true + ); + assert_split_line!("床前明月光，", "疑是地上霜。", true); + assert_split_line!("床前明月光。", "疑是地上霜。", true); + assert_split_line!("床前明月光；", "疑是地上霜。", true); + assert_split_line!( + "Wikipedia is (a free online encyclopedia).", + " that anyone can edit.", + true + ); + assert_split_line!( + "Wikipedia is a free online `encyclopedia,", + " that` anyone can edit.", + false + ); + assert_split_line!( + "Wikipedia is a free online ```encyclopedia,", + " that``` anyone can edit.", + false + ); + assert_split_line!( + "Wikipedia is a free online *encyclopedia,", + " that* anyone can edit.", + false + ); + assert_split_line!( + "Wikipedia is a free online **encyclopedia,", + " that** anyone can edit.", + false + ); + assert_split_line!( + "Wikipedia is a free online ~encyclopedia,", + " that~ anyone can edit.", + false + ); + assert_split_line!( + "Wikipedia is a free online ~~encyclopedia,", + " that~~ anyone can edit.", + false + ); + assert_split_line!( + "Wikipedia is a free online ``encyclopedia,", + " that`` anyone can edit.", + true + ); + assert_split_line!( + "Wikipedia is a free online \"encyclopedia,", + " that\" anyone can edit.", + false + ); + assert_split_line!( + "Wikipedia is a free online 'encyclopedia,", + " that' anyone can edit.", + false + ); + assert_split_line!( + "Wikipedia is a free online encyclopedia.", + "that anyone can edit.", + false + ); + } +} -- cgit v1.2.3