Summary
Query runs 2 operators; returns 100 rows in 0.08 ms. Main finding: Row estimate off by 10× or more: ss (estimated 1,666, actual 100 rows per loop, 17× over).
- Total time
- 84 µs
- Rows returned
- 100
- Rows examined
- 100
- Operators
- 2
Primary recommendation
Refresh statistics with ANALYZE TABLE ss. If a filter on a non-indexed column stays misestimated, add a histogram: on MySQL 8.0+, ANALYZE TABLE ss UPDATE HISTOGRAM ON <column> WITH 100 BUCKETS; on MariaDB, ANALYZE TABLE ss PERSISTENT FOR COLUMNS (<column>) INDEXES ().
Why does this help?
the optimizer picks join order and access methods from these estimates. InnoDB recalculates persistent statistics on its own only after about 10% of a table's rows change, and on MariaDB, under the default use_stat_tables=PREFERABLY_FOR_QUERIES, ANALYZE TABLE collects column statistics and histograms only when you add PERSISTENT FOR.
SQL
select testdb.ss.g AS g,testdb.ss.v AS v
from testdb.ss
where (testdb.ss.v > 4900)
Execution plan (flamegraph)
Click a plan operator to inspect details.
Warnings (1)
-
WARN
Row estimate off by 10× or more: ss (estimated 1,666, actual 100 rows per loop, 17× over)
Labeled:
Covering index [ss.g]
Suggestions (1)
-
MEDIUM
Refresh statistics with ANALYZE TABLE ss. If a filter on a non-indexed column stays misestimated, add a histogram: on MySQL 8.0+, ANALYZE TABLE ss UPDATE HISTOGRAM ON <column> WITH 100 BUCKETS; on MariaDB, ANALYZE TABLE ss PERSISTENT FOR COLUMNS (<column>) INDEXES ().
Why?
the optimizer picks join order and access methods from these estimates. InnoDB recalculates persistent statistics on its own only after about 10% of a table's rows change, and on MariaDB, under the default use_stat_tables=PREFERABLY_FOR_QUERIES, ANALYZE TABLE collects column statistics and histograms only when you add PERSISTENT FOR.
Agent-ready · token savings
Feeding this plan to an AI as the myflames digest instead of raw EXPLAIN JSON cuts the prompt by 1.2× (15.1% fewer tokens).
- Raw plan
- 392 tokens
- myflames digest
- 333 tokens
- Saved
- 59 tokens
- Per 1,000 queries
- ~$0.18 saved
Estimate (heuristic estimate (offline)). Prices as of 2026-05-26 (Sonnet 4.6 input). Generate the digest with myflames digest.
myteach — interactive algorithm catalog
Every operator shown above has a hands-on lesson with a cost-model slider and an animated walk-through. The catalog is the central index — bookmark it once, revisit per query:
Raw sidecar (JSON)
{
"$schema": "https://myflames.dev/schemas/sidecar-v1.json",
"schema_version": "1.3",
"generated_at": "2026-10-15T19:53:28Z",
"myflames_version": "2.3.0",
"source": {
"type": "file"
},
"plan_summary": {
"total_time_ms": 0.084,
"rows_sent": 100,
"rows_examined_estimate": 100,
"operator_count": 2,
"max_depth": 2
},
"optimizer_switches": [
{
"name": "skip_scan",
"value": "on",
"explanation": "Uses a composite index even though the query does not filter on the leading column, by probing each distinct leading-column value. Only pays off when the leading column has very low cardinality.",
"node_labels": [
"Covering index [ss.g]"
]
}
],
"warnings": [
{
"severity": "warn",
"category": "misestimate",
"text": "Row estimate off by 10× or more: ss (estimated 1,666, actual 100 rows per loop, 17× over)",
"source": "plan",
"node_labels": [
"Covering index [ss.g]"
]
}
],
"suggestions": [
{
"severity": "medium",
"category": "statistics",
"action": "Refresh statistics with ANALYZE TABLE ss. If a filter on a non-indexed column stays misestimated, add a histogram: on MySQL 8.0+, ANALYZE TABLE ss UPDATE HISTOGRAM ON <column> WITH 100 BUCKETS; on MariaDB, ANALYZE TABLE ss PERSISTENT FOR COLUMNS (<column>) INDEXES ().",
"source": "plan",
"why": "the optimizer picks join order and access methods from these estimates. InnoDB recalculates persistent statistics on its own only after about 10% of a table's rows change, and on MariaDB, under the default use_stat_tables=PREFERABLY_FOR_QUERIES, ANALYZE TABLE collects column statistics and histograms only when you add PERSISTENT FOR."
}
],
"executive_summary": "Query runs 2 operators; returns 100 rows in 0.08 ms. Main finding: Row estimate off by 10× or more: ss (estimated 1,666, actual 100 rows per loop, 17× over).",
"plan_tree": {
"node_id": "n:73b39ab87252",
"short_label": "Filter: ((ss.v > 4900))",
"folded_label": "FILTER ((ss.v > 4900)) starts=1 rows=100",
"children": [
{
"node_id": "n:df672e70bc82",
"short_label": "Covering index [ss.g]",
"folded_label": "COVERING INDEX [ss.g] starts=1 rows=100",
"children": []
}
]
},
"query": {
"raw": "/* select#1 */ select `testdb`.`ss`.`g` AS `g`,`testdb`.`ss`.`v` AS `v` from `testdb`.`ss` where (`testdb`.`ss`.`v` > 4900)",
"beautified": "select testdb.ss.g AS g,testdb.ss.v AS v\nfrom testdb.ss\nwhere (testdb.ss.v > 4900)"
},
"primary_action": {
"ref": "suggestions[0]"
},
"teach_hooks": [
{
"lesson": "filter",
"match": {
"folded_label": "FILTER ((ss.v > 4900)) starts=1 rows=100",
"short_label": "Filter: ((ss.v > 4900))"
},
"controls": {
"input_rows": 100,
"selectivity": 6.0
},
"note": "Filter: (ss.v > 4900)",
"query_sql": "/* select#1 */ select `testdb`.`ss`.`g` AS `g`,`testdb`.`ss`.`v` AS `v` from `testdb`.`ss` where (`testdb`.`ss`.`v` > 4900)"
},
{
"lesson": "skip_scan",
"match": {
"folded_label": "COVERING INDEX [ss.g] starts=1 rows=100",
"short_label": "Covering index [ss.g]"
},
"controls": {
"table_rows": 100,
"ndv_leading": 5,
"selectivity": 10
},
"note": "Covering index skip scan on ss using g over 4900 < v (ss)",
"query_sql": "/* select#1 */ select `testdb`.`ss`.`g` AS `g`,`testdb`.`ss`.`v` AS `v` from `testdb`.`ss` where (`testdb`.`ss`.`v` > 4900)"
}
],
"operator_complexities": [
{
"node_id": "n:df672e70bc82",
"folded_label": "COVERING INDEX [ss.g] starts=1 rows=100",
"short_label": "Covering index [ss.g]",
"complexity": {
"big_o": "O(n)",
"short": "n",
"severity": "good",
"rationale": "Covering index scan: walks the whole index in key order but never touches the clustered tree — all selected columns fit on the index leaves.",
"confidence": "exact",
"learn_more": "covering_index"
}
}
]
}