Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Return int32 for integer type date part #13466

Merged
merged 8 commits into from
Nov 22, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
127 changes: 99 additions & 28 deletions datafusion/functions/src/datetime/date_part.rs
Original file line number Diff line number Diff line change
Expand Up @@ -19,12 +19,11 @@ use std::any::Any;
use std::str::FromStr;
use std::sync::{Arc, OnceLock};

use arrow::array::{Array, ArrayRef, Float64Array};
use arrow::array::{Array, ArrayRef, Float64Array, Int32Array};
use arrow::compute::kernels::cast_utils::IntervalUnit;
use arrow::compute::{binary, cast, date_part, DatePart};
use arrow::compute::{binary, date_part, DatePart};
use arrow::datatypes::DataType::{
Date32, Date64, Duration, Float64, Interval, Time32, Time64, Timestamp, Utf8,
Utf8View,
Date32, Date64, Duration, Interval, Time32, Time64, Timestamp, Utf8, Utf8View,
};
use arrow::datatypes::IntervalUnit::{DayTime, MonthDayNano, YearMonth};
use arrow::datatypes::TimeUnit::{Microsecond, Millisecond, Nanosecond, Second};
Expand All @@ -36,11 +35,12 @@ use datafusion_common::cast::{
as_timestamp_microsecond_array, as_timestamp_millisecond_array,
as_timestamp_nanosecond_array, as_timestamp_second_array,
};
use datafusion_common::{exec_err, Result, ScalarValue};
use datafusion_common::{exec_err, internal_err, ExprSchema, Result, ScalarValue};
use datafusion_expr::scalar_doc_sections::DOC_SECTION_DATETIME;
use datafusion_expr::TypeSignature::Exact;
use datafusion_expr::{
ColumnarValue, Documentation, ScalarUDFImpl, Signature, Volatility, TIMEZONE_WILDCARD,
ColumnarValue, Documentation, Expr, ScalarUDFImpl, Signature, Volatility,
TIMEZONE_WILDCARD,
};

#[derive(Debug)]
Expand Down Expand Up @@ -148,7 +148,21 @@ impl ScalarUDFImpl for DatePartFunc {
}

fn return_type(&self, _arg_types: &[DataType]) -> Result<DataType> {
Ok(Float64)
internal_err!("return_type_from_exprs shoud be called instead")
}

fn return_type_from_exprs(
&self,
args: &[Expr],
_schema: &dyn ExprSchema,
_arg_types: &[DataType],
) -> Result<DataType> {
match &args[0] {
Expr::Literal(ScalarValue::Utf8(Some(part))) if is_epoch(part) => {
Ok(DataType::Float64)
}
_ => Ok(DataType::Int32),
}
}

fn invoke(&self, args: &[ColumnarValue]) -> Result<ColumnarValue> {
Expand All @@ -174,35 +188,31 @@ impl ScalarUDFImpl for DatePartFunc {
ColumnarValue::Scalar(scalar) => scalar.to_array()?,
};

// to remove quotes at most 2 characters
let part_trim = part.trim_matches(|c| c == '\'' || c == '\"');
if ![2, 0].contains(&(part.len() - part_trim.len())) {
return exec_err!("Date part '{part}' not supported");
}
let part_trim = part_normalization(part);

// using IntervalUnit here means we hand off all the work of supporting plurals (like "seconds")
// and synonyms ( like "ms,msec,msecond,millisecond") to Arrow
let arr = if let Ok(interval_unit) = IntervalUnit::from_str(part_trim) {
match interval_unit {
IntervalUnit::Year => date_part_f64(array.as_ref(), DatePart::Year)?,
IntervalUnit::Month => date_part_f64(array.as_ref(), DatePart::Month)?,
IntervalUnit::Week => date_part_f64(array.as_ref(), DatePart::Week)?,
IntervalUnit::Day => date_part_f64(array.as_ref(), DatePart::Day)?,
IntervalUnit::Hour => date_part_f64(array.as_ref(), DatePart::Hour)?,
IntervalUnit::Minute => date_part_f64(array.as_ref(), DatePart::Minute)?,
IntervalUnit::Second => seconds(array.as_ref(), Second)?,
IntervalUnit::Millisecond => seconds(array.as_ref(), Millisecond)?,
IntervalUnit::Microsecond => seconds(array.as_ref(), Microsecond)?,
IntervalUnit::Nanosecond => seconds(array.as_ref(), Nanosecond)?,
IntervalUnit::Year => date_part(array.as_ref(), DatePart::Year)?,
IntervalUnit::Month => date_part(array.as_ref(), DatePart::Month)?,
IntervalUnit::Week => date_part(array.as_ref(), DatePart::Week)?,
IntervalUnit::Day => date_part(array.as_ref(), DatePart::Day)?,
IntervalUnit::Hour => date_part(array.as_ref(), DatePart::Hour)?,
IntervalUnit::Minute => date_part(array.as_ref(), DatePart::Minute)?,
IntervalUnit::Second => seconds_as_i32(array.as_ref(), Second)?,
IntervalUnit::Millisecond => seconds_as_i32(array.as_ref(), Millisecond)?,
IntervalUnit::Microsecond => seconds_as_i32(array.as_ref(), Microsecond)?,
IntervalUnit::Nanosecond => seconds_as_i32(array.as_ref(), Nanosecond)?,
// century and decade are not supported by `DatePart`, although they are supported in postgres
_ => return exec_err!("Date part '{part}' not supported"),
}
} else {
// special cases that can be extracted (in postgres) but are not interval units
match part_trim.to_lowercase().as_str() {
"qtr" | "quarter" => date_part_f64(array.as_ref(), DatePart::Quarter)?,
"doy" => date_part_f64(array.as_ref(), DatePart::DayOfYear)?,
"dow" => date_part_f64(array.as_ref(), DatePart::DayOfWeekSunday0)?,
"qtr" | "quarter" => date_part(array.as_ref(), DatePart::Quarter)?,
"doy" => date_part(array.as_ref(), DatePart::DayOfYear)?,
"dow" => date_part(array.as_ref(), DatePart::DayOfWeekSunday0)?,
"epoch" => epoch(array.as_ref())?,
_ => return exec_err!("Date part '{part}' not supported"),
}
Expand All @@ -223,6 +233,18 @@ impl ScalarUDFImpl for DatePartFunc {
}
}

fn is_epoch(part: &str) -> bool {
let part = part_normalization(part);
matches!(part.to_lowercase().as_str(), "epoch")
}

// Try to remove quote if exist, if the quote is invalid, return original string and let the downstream function handle the error
fn part_normalization(part: &str) -> &str {
part.strip_prefix(|c| c == '\'' || c == '\"')
.and_then(|s| s.strip_suffix(|c| c == '\'' || c == '\"'))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why would we want to support unmatched quotes like '..."?

the date_part argument shouldn't contain any quotes, why would we want to support quotes at all?

i see this is pre-existing, so okay to keep, but still wonder why we're doing this, doesn't feel right

.unwrap_or(part)
}

static DOCUMENTATION: OnceLock<Documentation> = OnceLock::new();

fn get_date_part_doc() -> &'static Documentation {
Expand Down Expand Up @@ -261,14 +283,63 @@ fn get_date_part_doc() -> &'static Documentation {
})
}

/// Invoke [`date_part`] and cast the result to Float64
fn date_part_f64(array: &dyn Array, part: DatePart) -> Result<ArrayRef> {
Ok(cast(date_part(array, part)?.as_ref(), &Float64)?)
/// Invoke [`date_part`] on an `array` (e.g. Timestamp) and convert the
/// result to a total number of seconds, milliseconds, microseconds or
/// nanoseconds
fn seconds_as_i32(array: &dyn Array, unit: TimeUnit) -> Result<ArrayRef> {
// Nanosecond is neither supported in Postgres nor DuckDB, to avoid to deal with overflow and precision issue we don't support nanosecond
if unit == Nanosecond {
return internal_err!("unit {unit:?} not supported");
}

let conversion_factor = match unit {
Copy link
Contributor

@comphead comphead Nov 20, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we should probably factor out the common part for this conversion in future. its too many hardcoded conversions between second fraction and corresponding long value in the project.

Second => 1_000_000_000,
Millisecond => 1_000_000,
Microsecond => 1_000,
Nanosecond => 1,
};

let second_factor = match unit {
Second => 1,
Millisecond => 1_000,
Microsecond => 1_000_000,
Nanosecond => 1_000_000_000,
};

let secs = date_part(array, DatePart::Second)?;
// This assumes array is primitive and not a dictionary
let secs = as_int32_array(secs.as_ref())?;
let subsecs = date_part(array, DatePart::Nanosecond)?;
let subsecs = as_int32_array(subsecs.as_ref())?;

// Special case where there are no nulls.
if subsecs.null_count() == 0 {
let r: Int32Array = binary(secs, subsecs, |secs, subsecs| {
secs * second_factor + (subsecs % 1_000_000_000) / conversion_factor
})?;
Ok(Arc::new(r))
} else {
// Nulls in secs are preserved, nulls in subsecs are treated as zero to account for the case
// where the number of nanoseconds overflows.
let r: Int32Array = secs
.iter()
.zip(subsecs)
.map(|(secs, subsecs)| {
secs.map(|secs| {
let subsecs = subsecs.unwrap_or(0);
secs * second_factor + (subsecs % 1_000_000_000) / conversion_factor
})
})
.collect();
Ok(Arc::new(r))
}
}

/// Invoke [`date_part`] on an `array` (e.g. Timestamp) and convert the
/// result to a total number of seconds, milliseconds, microseconds or
/// nanoseconds
///
/// Given epoch return f64, this is a duplicated function to optimize for f64 type
fn seconds(array: &dyn Array, unit: TimeUnit) -> Result<ArrayRef> {
let sf = match unit {
Second => 1_f64,
Expand Down
2 changes: 1 addition & 1 deletion datafusion/sqllogictest/test_files/clickbench.slt
Original file line number Diff line number Diff line change
Expand Up @@ -136,7 +136,7 @@ SELECT "UserID", "SearchPhrase", COUNT(*) FROM hits GROUP BY "UserID", "SearchPh
519640690937130534 (empty) 2
7418527520126366595 (empty) 1

query IRTI rowsort
query IITI rowsort
SELECT "UserID", extract(minute FROM to_timestamp_seconds("EventTime")) AS m, "SearchPhrase", COUNT(*) FROM hits GROUP BY "UserID", m, "SearchPhrase" ORDER BY COUNT(*) DESC LIMIT 10;
----
-2461439046089301801 18 (empty) 1
Expand Down
Loading