Partition pruning fails for queries with newlines in SQL #146

Thank you for the detailed bug report and the fix, @khalid244! You're absolutely correct.

Confirmed Bug

The whereClausePattern regex uses .+? which does NOT match newlines in Go's regex engine. This causes partition pruning to fail for any multi-line SQL query.

Current code (partition_pruner.go:116):

whereClausePattern = regexp.MustCompile(`(?i)\bWHERE\b\s+(.+?)(?:\bGROUP BY\b|\bORDER BY\b|\bLIMIT\b|$)`)

The .+? pattern stops at newlines, so when your query has:

SELECT region, COUNT(*)
FROM metrics
WHERE time >= '2026-01-21T07:00:00Z' AND time < '2026-01-21T08:00:00Z'
GROUP BY region

The regex fails to capture the WHERE clause content, ExtractTimeRange() returns nil, and partition pruning is disabled entirely.

Your Fix is Correct

Changing .+? to [\s\S]+? is the right solution:

whereClausePattern = regexp.MustCompile(`(?i)\bWHERE\b\s+([\s\S]+?)(?:\bGROUP BY\b|\bORDER BY\b|\bLIMIT\b|$)`)

The [\s\S] pattern explicitly matches any character including newlines, which fixes the issue.

Would you like to submit a PR?

This is a simple one-line fix that would be perfect for a PR! If you'd like to submit it:

1. Change line 116 in internal/pruning/partition_pruner.go:

   whereClausePattern = regexp.MustCompile(`(?i)\bWHERE\b\s+([\s\S]+?)(?:\bGROUP BY\b|\bORDER BY\b|\bLIMIT\b|$)`)

2. Add a test case in internal/pruning/partition_pruner_test.go to prevent regression:

   func TestExtractTimeRange_MultilineQuery(t *testing.T) {
       sql := `SELECT region, COUNT(*)
   FROM metrics
   WHERE time >= '2026-01-21T07:00:00Z' AND time < '2026-01-21T08:00:00Z'
   GROUP BY region`
   
       pruner := NewPartitionPruner(config.PartitionPrunerConfig{Enabled: true}, nil, zerolog.Nop())
       timeRange := pruner.ExtractTimeRange(sql)
   
       require.NotNil(t, timeRange, "Should extract time range from multi-line query")
       assert.Equal(t, "2026-01-21T07:00:00Z", timeRange.Start.Format(time.RFC3339))
       assert.Equal(t, "2026-01-21T08:00:00Z", timeRange.End.Format(time.RFC3339))
   }

3. Run tests to verify no regressions:

   go test ./internal/pruning/...

Alternatively, I can apply this fix directly to the codebase if you prefer. Let me know!

Impact

This fix will:

• Enable partition pruning for multi-line SQL queries (Grafana, BI tools, formatted queries)
• Reduce S3 costs by avoiding full table scans
• Improve query performance significantly
• No breaking changes or regressions

Great catch on this bug! Multi-line queries are very common in practice (Grafana dashboards, SQL editors, formatted scripts), so this fix will have significant impact.

Done #148