Faster time parsing (~93% faster) #3860
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,42 @@ | ||
| // Licensed to the Apache Software Foundation (ASF) under one | ||
| // or more contributor license agreements. See the NOTICE file | ||
| // distributed with this work for additional information | ||
| // regarding copyright ownership. The ASF licenses this file | ||
| // to you under the Apache License, Version 2.0 (the | ||
| // "License"); you may not use this file except in compliance | ||
| // with the License. You may obtain a copy of the License at | ||
| // | ||
| // http://www.apache.org/licenses/LICENSE-2.0 | ||
| // | ||
| // Unless required by applicable law or agreed to in writing, | ||
| // software distributed under the License is distributed on an | ||
| // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY | ||
| // KIND, either express or implied. See the License for the | ||
| // specific language governing permissions and limitations | ||
| // under the License. | ||
|
|
||
| use arrow_cast::parse::string_to_time_nanoseconds; | ||
| use criterion::*; | ||
|
|
||
| fn criterion_benchmark(c: &mut Criterion) { | ||
| let timestamps = [ | ||
| "9:50", | ||
| "09:50", | ||
| "09:50 PM", | ||
| "9:50:12 AM", | ||
| "09:50:12 PM", | ||
| "09:50:12.123456789", | ||
| "9:50:12.123456789", | ||
| "09:50:12.123456789 PM", | ||
| ]; | ||
|
|
||
| for timestamp in timestamps { | ||
| let t = black_box(timestamp); | ||
| c.bench_function(t, |b| { | ||
| b.iter(|| string_to_time_nanoseconds(t).unwrap()); | ||
| }); | ||
| } | ||
| } | ||
|
|
||
| criterion_group!(benches, criterion_benchmark); | ||
| criterion_main!(benches); |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -23,11 +23,12 @@ use arrow_schema::ArrowError; | |
| use chrono::prelude::*; | ||
| use std::str::FromStr; | ||
|
|
||
| /// Parse nanoseconds from the first `N` values in digits, subtracting the offset `O` | ||
| #[inline] | ||
| fn parse_nanos<const N: usize, const O: u8>(digits: &[u8]) -> u32 { | ||
| digits[..N] | ||
| .iter() | ||
| .fold(0_u32, |acc, v| acc * 10 + v.wrapping_sub(O) as u32) | ||
| * 10_u32.pow((9 - N) as _) | ||
| } | ||
|
|
||
|
|
@@ -102,15 +103,15 @@ impl TimestampParser { | |
| let digits = (self.mask >> 20).trailing_ones(); | ||
| let nanos = match digits { | ||
| 0 => return None, | ||
| 1 => parse_nanos::<1, 0>(&self.digits[20..21]), | ||
| 2 => parse_nanos::<2, 0>(&self.digits[20..22]), | ||
| 3 => parse_nanos::<3, 0>(&self.digits[20..23]), | ||
| 4 => parse_nanos::<4, 0>(&self.digits[20..24]), | ||
| 5 => parse_nanos::<5, 0>(&self.digits[20..25]), | ||
| 6 => parse_nanos::<6, 0>(&self.digits[20..26]), | ||
| 7 => parse_nanos::<7, 0>(&self.digits[20..27]), | ||
| 8 => parse_nanos::<8, 0>(&self.digits[20..28]), | ||
| _ => parse_nanos::<9, 0>(&self.digits[20..29]), | ||
| }; | ||
| Some((time.with_nanosecond(nanos)?, 20 + digits as usize)) | ||
| } | ||
|
|
@@ -291,79 +292,120 @@ fn to_timestamp_nanos(dt: NaiveDateTime) -> Result<i64, ArrowError> { | |
| /// This function does not support parsing strings with a timezone | ||
| /// or offset specified, as it considers only time since midnight. | ||
| pub fn string_to_time_nanoseconds(s: &str) -> Result<i64, ArrowError> { | ||
| let nt = string_to_time(s).ok_or_else(|| { | ||
| ArrowError::ParseError(format!("Failed to parse \'{s}\' as time")) | ||
| })?; | ||
| Ok(nt.num_seconds_from_midnight() as i64 * 1_000_000_000 + nt.nanosecond() as i64) | ||
| } | ||
|
|
||
| fn string_to_time(s: &str) -> Option<NaiveTime> { | ||
| let bytes = s.as_bytes(); | ||
| if bytes.len() < 4 { | ||
| return None; | ||
| } | ||
|
|
||
| let (am, bytes) = match bytes.get(bytes.len() - 3..) { | ||
| Some(b" AM" | b" am" | b" Am" | b" aM") => { | ||
| (Some(true), &bytes[..bytes.len() - 3]) | ||
| } | ||
| Some(b" PM" | b" pm" | b" pM" | b" Pm") => { | ||
| (Some(false), &bytes[..bytes.len() - 3]) | ||
| } | ||
| _ => (None, bytes), | ||
| }; | ||
|
|
||
| if bytes.len() < 4 { | ||
| return None; | ||
| } | ||
|
|
||
| let mut digits = [b'0'; 6]; | ||
|
|
||
| // Extract hour | ||
| let bytes = match (bytes[1], bytes[2]) { | ||
| (b':', _) => { | ||
| digits[1] = bytes[0]; | ||
| &bytes[2..] | ||
| } | ||
| (_, b':') => { | ||
| digits[0] = bytes[0]; | ||
| digits[1] = bytes[1]; | ||
| &bytes[3..] | ||
| } | ||
| _ => return None, | ||
| }; | ||
|
|
||
| if bytes.len() < 2 { | ||
| return None; // Minutes required | ||
| } | ||
|
|
||
| // Extract minutes | ||
| digits[2] = bytes[0]; | ||
| digits[3] = bytes[1]; | ||
|
|
||
| let nanoseconds = match bytes.get(2) { | ||
| Some(b':') => { | ||
| if bytes.len() < 5 { | ||
| return None; | ||
| } | ||
|
|
||
| // Extract seconds | ||
| digits[4] = bytes[3]; | ||
| digits[5] = bytes[4]; | ||
|
|
||
| // Extract sub-seconds if any | ||
| match bytes.get(5) { | ||
| Some(b'.') => { | ||
| let decimal = &bytes[6..]; | ||
| if decimal.iter().any(|x| !x.is_ascii_digit()) { | ||
| return None; | ||
| } | ||
| match decimal.len() { | ||
| 0 => return None, | ||
| 1 => parse_nanos::<1, b'0'>(decimal), | ||
| 2 => parse_nanos::<2, b'0'>(decimal), | ||
| 3 => parse_nanos::<3, b'0'>(decimal), | ||
| 4 => parse_nanos::<4, b'0'>(decimal), | ||
| 5 => parse_nanos::<5, b'0'>(decimal), | ||
| 6 => parse_nanos::<6, b'0'>(decimal), | ||
| 7 => parse_nanos::<7, b'0'>(decimal), | ||
| 8 => parse_nanos::<8, b'0'>(decimal), | ||
| _ => parse_nanos::<9, b'0'>(decimal), | ||
| } | ||
| } | ||
| Some(_) => return None, | ||
| None => 0, | ||
| } | ||
| } | ||
| Some(_) => return None, | ||
| None => 0, | ||
| }; | ||
|
|
||
| digits.iter_mut().for_each(|x| *x = x.wrapping_sub(b'0')); | ||
| if digits.iter().any(|x| *x > 9) { | ||
| return None; | ||
| } | ||
|
|
||
| let hour = match (digits[0] * 10 + digits[1], am) { | ||
| (12, Some(true)) => 0, // 12:00 AM -> 00:00 | ||
| (h @ 1..=11, Some(true)) => h, // 1:00 AM -> 01:00 | ||
| (12, Some(false)) => 12, // 12:00 PM -> 12:00 | ||
| (h @ 1..=11, Some(false)) => h + 12, // 1:00 PM -> 13:00 | ||
| (_, Some(_)) => return None, | ||
| (h, None) => h, | ||
| }; | ||
|
|
||
| // Handle leap second | ||
| let (second, nanoseconds) = match digits[4] * 10 + digits[5] { | ||
| 60 => (59, nanoseconds + 1_000_000_000), | ||
| s => (s, nanoseconds), | ||
| }; | ||
|
|
||
| NaiveTime::from_hms_nano_opt( | ||
| hour as _, | ||
| (digits[2] * 10 + digits[3]) as _, | ||
| second as _, | ||
| nanoseconds, | ||
| ) | ||
| } | ||
|
|
||
| /// Specialized parsing implementations | ||
|
|
@@ -892,6 +934,13 @@ mod tests { | |
| use arrow_array::timezone::Tz; | ||
| use arrow_buffer::i256; | ||
|
|
||
| #[test] | ||
| fn test_parse_nanos() { | ||
| assert_eq!(parse_nanos::<3, 0>(&[1, 2, 3]), 123_000_000); | ||
| assert_eq!(parse_nanos::<5, 0>(&[1, 2, 3, 4, 5]), 123_450_000); | ||
| assert_eq!(parse_nanos::<6, b'0'>(b"123456"), 123_456_000); | ||
| } | ||
|
|
||
| #[test] | ||
| fn string_to_timestamp_timezone() { | ||
| // Explicit timezone | ||
|
|
@@ -1382,6 +1431,74 @@ mod tests { | |
| ); | ||
| } | ||
|
|
||
| #[test] | ||
| fn test_string_to_time_invalid() { | ||
| let cases = [ | ||
| "25:00", | ||
|
There was a problem hiding this comment. I recommend also testing space and leading colon There was a problem hiding this comment. Also, very long subseconds: There was a problem hiding this comment. Also intermediate non digits: |
||
| "9:00:", | ||
| "009:00", | ||
| "09:0:00", | ||
| "25:00:00", | ||
| "13:00 AM", | ||
| "13:00 PM", | ||
| "12:00. AM", | ||
| "09:0:00", | ||
| "09:01:0", | ||
| "09:01:1", | ||
| "9:1:0", | ||
| "09:01:0", | ||
| "1:00.123", | ||
| "1:00:00.123f", | ||
| " 9:00:00", | ||
| ":09:00", | ||
| "T9:00:00", | ||
| "AM", | ||
| ]; | ||
| for case in cases { | ||
| assert!(string_to_time(case).is_none(), "{case}"); | ||
| } | ||
| } | ||
|
|
||
| #[test] | ||
| fn test_string_to_time_chrono() { | ||
| let cases = [ | ||
| ("1:00", "%H:%M"), | ||
| ("12:00", "%H:%M"), | ||
| ("13:00", "%H:%M"), | ||
| ("24:00", "%H:%M"), | ||
| ("1:00:00", "%H:%M:%S"), | ||
| ("12:00:30", "%H:%M:%S"), | ||
| ("13:00:59", "%H:%M:%S"), | ||
| ("24:00:60", "%H:%M:%S"), | ||
| ("09:00:00", "%H:%M:%S%.f"), | ||
| ("0:00:30.123456", "%H:%M:%S%.f"), | ||
| ("0:00 AM", "%I:%M %P"), | ||
| ("1:00 AM", "%I:%M %P"), | ||
| ("12:00 AM", "%I:%M %P"), | ||
| ("13:00 AM", "%I:%M %P"), | ||
| ("0:00 PM", "%I:%M %P"), | ||
| ("1:00 PM", "%I:%M %P"), | ||
| ("12:00 PM", "%I:%M %P"), | ||
| ("13:00 PM", "%I:%M %P"), | ||
| ("1:00 pM", "%I:%M %P"), | ||
| ("1:00 Pm", "%I:%M %P"), | ||
| ("1:00 aM", "%I:%M %P"), | ||
| ("1:00 Am", "%I:%M %P"), | ||
| ("1:00:30.123456 PM", "%I:%M:%S%.f %P"), | ||
| ("1:00:30.123456789 PM", "%I:%M:%S%.f %P"), | ||
| ("1:00:30.123456789123 PM", "%I:%M:%S%.f %P"), | ||
|
There was a problem hiding this comment. Given the experience with #3859, I think it would be valuable to include a time that has 4 and 7 significant digit (aka not a multiple of three) Perhaps? |
||
| ("1:00:30.1234 PM", "%I:%M:%S%.f %P"), | ||
| ("1:00:30.123456 PM", "%I:%M:%S%.f %P"), | ||
| ("1:00:30.123456789123456789 PM", "%I:%M:%S%.f %P"), | ||
| ("1:00:30.12F456 PM", "%I:%M:%S%.f %P"), | ||
| ]; | ||
| for (s, format) in cases { | ||
| let chrono = NaiveTime::parse_from_str(s, format).ok(); | ||
| let custom = string_to_time(s); | ||
| assert_eq!(chrono, custom, "{s}"); | ||
| } | ||
| } | ||
|
|
||
| #[test] | ||
| fn test_parse_interval() { | ||
| assert_eq!( | ||
|
|
||
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
👍