mirror of
https://github.com/go-vikunja/vikunja.git
synced 2026-11-16 14:42:32 -06:00
perf(tasks): only use DISTINCT when a join can multiply rows
The task search always emitted SELECT DISTINCT tasks.*, but the only join that can produce duplicate rows is task_buckets without a project view to pin it — task_positions and task_buckets are both unique on (task_id, project_view_id), and every other condition filters through a subquery. DISTINCT forced a sort+unique over every matching row before the LIMIT could apply, so `ORDER BY due_date ... LIMIT n` on the home overview could never walk an index. On MySQL the longtext description also pushed it into an on-disk temp table. Gate it, and the count query along with it. Selecting task_positions.position was only needed to satisfy Postgres' rule that ORDER BY expressions appear in a DISTINCT select list, so that goes too.
This commit is contained in:
@@ -583,9 +583,17 @@ func (d *dbTaskSearcher) Search(opts *taskSearchOptions) (tasks []*Task, totalCo
|
||||
return nil, 0, err
|
||||
}
|
||||
|
||||
var distinct = "tasks.*"
|
||||
if strings.Contains(orderby, "task_positions.") {
|
||||
distinct += ", task_positions.position"
|
||||
// The only join that can multiply rows is task_buckets without a view to pin
|
||||
// it — task_positions and task_buckets are both unique on (task_id,
|
||||
// project_view_id), everything else filters through subqueries. DISTINCT
|
||||
// otherwise costs a full sort+unique of every matching row before the LIMIT,
|
||||
// which is what stops `ORDER BY due_date … LIMIT n` from walking an index.
|
||||
needsDistinct := joinTaskBuckets && opts.projectViewID == 0
|
||||
|
||||
var selectCols = "tasks.*"
|
||||
if needsDistinct && strings.Contains(orderby, "task_positions.") {
|
||||
// Postgres requires ORDER BY expressions to be in the DISTINCT select list.
|
||||
selectCols += ", task_positions.position"
|
||||
}
|
||||
|
||||
if expandSubtasks {
|
||||
@@ -593,13 +601,19 @@ func (d *dbTaskSearcher) Search(opts *taskSearchOptions) (tasks []*Task, totalCo
|
||||
}
|
||||
|
||||
query := d.s.Where(cond)
|
||||
if rankByRelevance {
|
||||
switch {
|
||||
case rankByRelevance:
|
||||
// Select() passes the raw column list through untouched while Distinct()
|
||||
// (no args) still emits DISTINCT. Distinct("tasks.*, pdb.score(tasks.id)")
|
||||
// would quote-corrupt the function call into "pdb"."score(tasks"."id)".
|
||||
query = query.Select(distinct + ", pdb.score(tasks.id)").Distinct()
|
||||
} else {
|
||||
query = query.Distinct(distinct)
|
||||
query = query.Select(selectCols + ", pdb.score(tasks.id)")
|
||||
if needsDistinct {
|
||||
query = query.Distinct()
|
||||
}
|
||||
case needsDistinct:
|
||||
query = query.Distinct(selectCols)
|
||||
default:
|
||||
query = query.Select(selectCols)
|
||||
}
|
||||
if limit > 0 {
|
||||
query = query.Limit(limit, start)
|
||||
@@ -690,8 +704,12 @@ func (d *dbTaskSearcher) Search(opts *taskSearchOptions) (tasks []*Task, totalCo
|
||||
queryCount = queryCount.Join("LEFT", "task_buckets", joinCond)
|
||||
}
|
||||
}
|
||||
countCol := "count(tasks.id)"
|
||||
if needsDistinct {
|
||||
countCol = "count(DISTINCT tasks.id)"
|
||||
}
|
||||
totalCount, err = queryCount.
|
||||
Select("count(DISTINCT tasks.id)").
|
||||
Select(countCol).
|
||||
Count(&Task{})
|
||||
if err != nil {
|
||||
sql, vals := queryCount.LastSQL()
|
||||
|
||||
Reference in New Issue
Block a user