findepi commented on code in PR #12978:
URL: https://github.com/apache/datafusion/pull/12978#discussion_r1826283217
##########
datafusion/core/src/physical_optimizer/pruning.rs:
##########
@@ -1610,6 +1629,126 @@ fn build_statistics_expr(
Ok(statistics_expr)
}
+fn build_like_match(
+ expr_builder: &mut PruningExpressionBuilder,
+) -> Option<Arc<dyn PhysicalExpr>> {
+ // column LIKE literal => (min, max) LIKE literal split at % => min <=
split literal && split literal <= max
+ // column LIKE 'foo%' => min <= 'foo' && 'foo' <= max
+ // column LIKE '%foo' => min <= '' && '' <= max => true
+ // column LIKE '%foo%' => min <= '' && '' <= max => true
+ // column LIKE 'foo' => min <= 'foo' && 'foo' <= max
+
+ fn unpack_string(s: &ScalarValue) -> Option<&String> {
+ match s {
+ ScalarValue::Utf8(Some(s)) => Some(s),
+ ScalarValue::LargeUtf8(Some(s)) => Some(s),
+ ScalarValue::Utf8View(Some(s)) => Some(s),
+ ScalarValue::Dictionary(_, value) => unpack_string(value),
+ _ => None,
+ }
+ }
+
+ fn extract_string_literal(expr: &Arc<dyn PhysicalExpr>) -> Option<&String>
{
+ if let Some(lit) = expr.as_any().downcast_ref::<phys_expr::Literal>() {
+ let s = unpack_string(lit.value())?;
+ return Some(s);
+ }
+ None
+ }
+
+ // TODO Handle ILIKE perhaps by making the min lowercase and max uppercase
+ // this may involve building the physical expressions that call lower()
and upper()
+ let min_column_expr = expr_builder.min_column_expr().ok()?;
+ let max_column_expr = expr_builder.max_column_expr().ok()?;
+ let scalar_expr = expr_builder.scalar_expr();
+ // check that the scalar is a string literal
+ let s = extract_string_literal(scalar_expr)?;
+ // ANSI SQL specifies two wildcards: % and _. % matches zero or more
characters, _ matches exactly one character.
+ let first_wildcard_index = s.find(['%', '_']);
+ if first_wildcard_index == Some(0) {
+ // there's no filtering we could possibly do, return an error and have
this be handled by the unhandled hook
+ return None;
+ }
+ let (upper_bound, lower_bound) = if let Some(wildcard_index) =
first_wildcard_index {
+ let prefix = &s[..wildcard_index];
+ let upper_bound_lit =
Arc::new(phys_expr::Literal::new(ScalarValue::Utf8(Some(
+ increment_utf8(prefix)?,
+ ))));
+ let lower_bound_lit =
Arc::new(phys_expr::Literal::new(ScalarValue::Utf8(Some(
+ prefix.to_string(),
+ ))));
+ (upper_bound_lit, lower_bound_lit)
Review Comment:
nit: can you perhaps compute and return lower bound first?
##########
datafusion/core/src/physical_optimizer/pruning.rs:
##########
@@ -1610,6 +1629,124 @@ fn build_statistics_expr(
Ok(statistics_expr)
}
+fn build_like_match(
+ expr_builder: &mut PruningExpressionBuilder,
+) -> Option<Arc<dyn PhysicalExpr>> {
+ // column LIKE literal => (min, max) LIKE literal split at % => min <=
split literal && split literal <= max
+ // column LIKE 'foo%' => min <= 'foo' && 'foo' <= max
+ // column LIKE '%foo' => min <= '' && '' <= max => true
+ // column LIKE '%foo%' => min <= '' && '' <= max => true
+ // column LIKE 'foo' => min <= 'foo' && 'foo' <= max
Review Comment:
see https://github.com/apache/datafusion/pull/12978#discussion_r1819023885
https://github.com/apache/datafusion/pull/12978#discussion_r1823893452
there could be a function doc comment
```
/// Convert `column LIKE literal` where P is a constant prefix of the literal
/// to a range check on the column: `P <= column && column < P'`, where P'
is the
/// lowest string after all P* strings.
```
--
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]
