Skip to content

Commit

Permalink
Browse files Browse the repository at this point in the history
38036: opt: fix overflow causing zero row-count r=justinj a=justinj

We were subtracting two ints that could overflow and then casting the
result to a float64. There's an easy solution to avoid the overflow:
just cast each integer to a float64 *before* performing the subtraction.

I will backport this to 19.1 and 2.1.

Release note (bug fix): Previously, due to a bug when estimating result
set sizes in the optimizer, queries involving int ranges that were very
large could result in poor plans being generated.

Co-authored-by: Justin Jaffray <justin@cockroachlabs.com>
  • Loading branch information
craig[bot] and Justin Jaffray committed Jun 5, 2019
2 parents ad21052 + 87c2c1d commit 279dfa5
Show file tree
Hide file tree
Showing 2 changed files with 17 additions and 3 deletions.
7 changes: 4 additions & 3 deletions pkg/sql/opt/memo/statistics_builder.go
Original file line number Diff line number Diff line change
Expand Up @@ -2525,11 +2525,12 @@ func (sb *statisticsBuilder) updateDistinctCountsFromConstraint(
end := int(*endVal.(*tree.DInt))
// We assume that both start and end boundaries are inclusive. This
// should be the case for integer valued columns (due to normalization
// by constraint.PreferInclusive).
// by constraint.PreferInclusive). We must cast each end to a float
// *before* performing the subtraction to avoid overflow.
if c.Columns.Get(col).Ascending() {
distinctCount += float64(end - start)
distinctCount += float64(end) - float64(start)
} else {
distinctCount += float64(start - end)
distinctCount += float64(start) - float64(end)
}
} else {
// We can't determine the distinct count for this column. For example,
Expand Down
13 changes: 13 additions & 0 deletions pkg/sql/opt/memo/testdata/stats/scan
Original file line number Diff line number Diff line change
Expand Up @@ -509,3 +509,16 @@ SELECT * FROM empty
scan empty
├── columns: x:1(int)
└── stats: [rows=1]

# Regression test: previously, overflow when computing estimated distinct count
# here resulted in a row count of zero being estimated.
opt
SELECT x FROM a WHERE x >= -9223372036854775808 AND x <= 0 ORDER BY x LIMIT 10
----
scan a
├── columns: x:1(int!null)
├── constraint: /1: [/-9223372036854775808 - /0]
├── limit: 10
├── stats: [rows=10]
├── key: (1)
└── ordering: +1

0 comments on commit 279dfa5

Please sign in to comment.