Druva-D opened a new issue, #20266:
URL: https://github.com/apache/datafusion/issues/20266
### Describe the bug
The FormatStringFunc
(datafusion/spark/src/function/string/format_string.rs) correctly parses the
grouping separator flag (,) in format specifiers like %,.2f, setting
`ConversionSpecifier::grouping_separator =true`
However, the flag is only applied during integer formatting in
format_unsigned() (for %d). It is silently ignored in two methods:
1. format_float() (line ~1684) -- handles Float16, Float32, Float64 with
%f, %e, %g conversions. The method never checks self.grouping_separator.
2. format_decimal() (line ~2022) -- handles Decimal128, Decimal256 with
%f, %e, %g conversions. Same issue.
This is inconsistent with `java.util.Formatter` behavior, where %,.2f
applies thousands separators to the integer portion of floating-point output.
### To Reproduce
```rs
use datafusion::execution::SessionStateBuilder;
use datafusion::prelude::SessionContext;
use datafusion_spark::SessionStateBuilderSpark;
#[tokio::main]
async fn main() -> datafusion::common::Result<()> {
let state = SessionStateBuilder::new()
.with_default_features()
.with_spark_features()
.build();
let ctx = SessionContext::new_with_state(state);
let df = ctx
.sql("SELECT format_string('%,.2f', 1234567.89)")
.await?;
df.show().await?;
Ok(())
}
```
Output
```
+--------------------------------------------------+
| format_string(Utf8("%,.2f"),Float64(1234567.89)) |
+--------------------------------------------------+
| 1234567.89 |
+--------------------------------------------------+
+--------------------------------------------------+
| format_string(Utf8("%,.2f"),Float64(1234567.89)) |
+--------------------------------------------------+
| 1234567.89 |
+--------------------------------------------------+
```
### Expected behavior
Got expected behaviour from Java
```java
public class FormatTest {
public static void main(String[] args) {
// Float grouping separator
System.out.println("Float %,.2f: " + String.format("%,.2f",
1234567.89));
System.out.println("Float %.2f: " + String.format("%.2f",
1234567.89));
System.out.println("Float %,.0f: " + String.format("%,.0f",
1000.0));
// Decimal (BigDecimal) grouping separator
java.math.BigDecimal bd = new java.math.BigDecimal("1234567.89");
System.out.println("BigDec %,.2f: " + String.format("%,.2f", bd));
System.out.println("BigDec %.2f: " + String.format("%.2f", bd));
}
}
```
```
Float %,.2f: 1,234,567.89
Float %.2f: 1234567.89
Float %,.0f: 1,000
BigDec %,.2f: 1,234,567.89
BigDec %.2f: 1234567.89
```
### Additional context
Apply the same thousands-separator loop already used in format_unsigned() to
the integer part, then rejoin. This
needs to be added in two places:
- format_float(): after line ~1768 where number = format!("{abs:.prec$}",
...)
- format_decimal(): after the number is produced from the match
self.conversion_type block (~line 2095), before padding is applied
Reference: the grouping logic already exists in format_unsigned() (line
~1897) and can be reused.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
