Add support for TABLESAMPLE pipe operator
#1860
Conversation
src/ast/query.rs
Outdated
| @@ -2680,28 +2680,32 @@ pub enum PipeOperator { | |||
| full_table_exprs: Vec<ExprWithAliasAndOrderBy>, | |||
| group_by_expr: Vec<ExprWithAliasAndOrderBy>, | |||
| }, | |||
| /// Selects a random sample of rows from the input table. | |||
| /// Syntax: `|> TABLESAMPLE <method> (<size> {ROWS | PERCENT})` | |||
There was a problem hiding this comment.
This is the officially supported spec from the paper but doesn't cover everything that is technically supported in this PR. I'm wondering how to best deal with that (see the open question in the PR description).
There was a problem hiding this comment.
I think the doc comment doesn't necessarily need to spell out the full spec. usually its enough to give a rough example of what the statement looks like e.g.
Syntax: `|> TABLESAMPLE BERNOULLI(50)
There was a problem hiding this comment.
Thanks @hendrikmakait! The changes look reasonable to me overall, left a couple minor comments
src/ast/query.rs
Outdated
| } | ||
|
|
||
| impl fmt::Display for PipeOperator { | ||
| fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { | ||
| match self { | ||
| PipeOperator::Select { exprs } => { | ||
| write!(f, "SELECT {}", display_comma_separated(exprs.as_slice())) | ||
| write!(f, " SELECT {}", display_comma_separated(exprs.as_slice())) |
There was a problem hiding this comment.
Oh I think we'd ideally undo these changes, its better for the operator display to be standalone where possible (i.e. they shouldn't assume surrounding space formatting, rather it should be left up to the caller), it makes it easier to compose nodes in the AST when displaying
There was a problem hiding this comment.
Good point, I found a way to fix the formatting for TableSample in a different way that requires fewer changes and follows this general principle.
src/ast/query.rs
Outdated
| @@ -2680,28 +2680,32 @@ pub enum PipeOperator { | |||
| full_table_exprs: Vec<ExprWithAliasAndOrderBy>, | |||
| group_by_expr: Vec<ExprWithAliasAndOrderBy>, | |||
| }, | |||
| /// Selects a random sample of rows from the input table. | |||
| /// Syntax: `|> TABLESAMPLE <method> (<size> {ROWS | PERCENT})` | |||
There was a problem hiding this comment.
I think the doc comment doesn't necessarily need to spell out the full spec. usually its enough to give a rough example of what the statement looks like e.g.
Syntax: `|> TABLESAMPLE BERNOULLI(50)
| @@ -1559,7 +1559,7 @@ impl fmt::Display for TableSampleBucket { | |||
| } | |||
| impl fmt::Display for TableSample { | |||
| fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { | |||
| write!(f, " {}", self.modifier)?; | |||
| write!(f, "{}", self.modifier)?; | |||
There was a problem hiding this comment.
For better composition of AST building blocks, I remove the whitespace here and introduce it again for the structs that contain a TableSample.
|
@hendrikmakait could you take a look at the CI failures? |
|
@iffyio: CI should be fixed now |
|
I restarted the checks |
|
🚀 |
Part of #1758
This PR adds support for the
|> TABLESAMPLE ...pipe operator.Open question:
BigQuery's
TABLESAMPLEoperator does not support the full SQL standard spec. This PR adds support for a superset of the BigQuery syntax which, e.g., includes theREPEATABLEclause and reuses the existing code for parsingTABLESAMPLEoperators. Is this desired from the perspective of the project or is it preferred to stick closely to the BigQuery syntax? If so, I'm happy to change this.