MDEV-26519: Improved histograms: Make JSON parser efficient

Previous JSON parser was using an API which made the parsing
inefficient: the same JSON contents was parsed again and again.

Switch to using a lower-level parsing API which allows to do
parsing in an efficient way.

mysql-test/main/statistics_json.result

@@ -4263,54 +4263,79 @@ UPDATE mysql.column_stats
 SET histogram='["not-what-you-expect"]' WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: Root JSON element must be a JSON object at offset 0.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
+Warnings:
+Warning	4186	Failed to parse histogram for table test.t1_json: Root JSON element must be a JSON object at offset 1.
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":"not-histogram"}' WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: A JSON array expected at offset 0.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
+Warnings:
+Warning	4186	Failed to parse histogram for table test.t1_json: histogram_hb_v2 must contain an array at offset 35.
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":["not-a-bucket"]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: Object expected at offset 19.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
+Warnings:
+Warning	4186	Failed to parse histogram for table test.t1_json: Expected an object in the buckets array at offset 35.
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"no-expected-members":1}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: .start member must be present and be a scalar at offset 20.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
+Warnings:
+Warning	4186	Failed to parse histogram for table test.t1_json: "start" element not present at offset 45.
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"start":{}}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: .start member must be present and be a scalar at offset 20.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
+Warnings:
+Warning	4186	Failed to parse histogram for table test.t1_json: "size" element not present at offset 31.
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"start":"aaa", "size":"not-an-integer"}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: .size member must be present and be a scalar at offset 20.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
+Warnings:
+Warning	4186	Failed to parse histogram for table test.t1_json: "ndv" element not present at offset 60.
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"start":"aaa", "size":0.25}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: .ndv member must be present and be a scalar at offset 20.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
+Warnings:
+Warning	4186	Failed to parse histogram for table test.t1_json: "ndv" element not present at offset 48.
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"start":"aaa", "size":0.25, "ndv":1}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: .end must be present in the last bucket and only there at offset 0.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
 explain select * from t1_json limit 1;
-ERROR HY000: Failed to parse histogram: .end must be present in the last bucket and only there at offset 0.
+id	select_type	table	type	possible_keys	key	key_len	ref	rows	Extra
+1	SIMPLE	t1_json	ALL	NULL	NULL	NULL	NULL	10	
+Warnings:
+Warning	4186	Failed to parse histogram for table test.t1_json: Histogram must have at least one bucket at offset 21.
 create table t2 (
 city varchar(100)
 );

mysql-test/main/statistics_json.test

@@ -46,62 +46,53 @@ drop table ten;
 UPDATE mysql.column_stats 
 SET histogram='["not-what-you-expect"]' WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":"not-histogram"}' WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":["not-a-bucket"]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"no-expected-members":1}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"start":{}}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"start":"aaa", "size":"not-an-integer"}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"start":"aaa", "size":0.25}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[{"start":"aaa", "size":0.25, "ndv":1}]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 UPDATE mysql.column_stats 
 SET histogram='{"histogram_hb_v2":[]}'
 WHERE table_name='t1_json';
 FLUSH TABLES;
---error ER_JSON_HISTOGRAM_PARSE_FAILED
 explain select * from t1_json limit 1;
 
 --source include/have_sequence.inc

sql/opt_histogram_json.h

@@ -76,7 +76,8 @@ class Histogram_json_hb : public Histogram_base
 public:
   static constexpr const char* JSON_NAME="histogram_hb_v2";
 
-  bool parse(MEM_ROOT *mem_root, Field *field, Histogram_type type_arg,
+  bool parse(MEM_ROOT *mem_root, const char *db_name, const char *table_name,
+             Field *field, Histogram_type type_arg,
              const char *hist_data, size_t hist_data_len) override;
 
   void serialize(Field *field) override;
@@ -122,6 +123,9 @@ class Histogram_json_hb : public Histogram_base
   }
 
 private:
+  int parse_bucket(json_engine_t *je, Field *field, double *cumulative_size,
+                   bool *assigned_last_end, const char **err);
+
   double get_left_fract(int idx);
   std::string& get_end_value(int idx);
   int find_bucket(const Field *field, const uchar *lookup_val, bool *equal);

sql/share/errmsg-utf8.txt

@@ -8914,4 +8914,4 @@ ER_PARTITION_CONVERT_SUBPARTITIONED
 ER_PROVIDER_NOT_LOADED
         eng "MariaDB tried to use the %s, but its provider plugin is not loaded"
 ER_JSON_HISTOGRAM_PARSE_FAILED
-        eng "Failed to parse histogram: %s at offset %d."
+        eng "Failed to parse histogram for table %s.%s: %s at offset %d."

sql/sql_statistics.cc

@@ -1233,7 +1233,8 @@ class Column_stat: public Stat_table
       if (!(hist= create_histogram(mem_root, hist_type, NULL)))
         return NULL;
       Field *field= table->field[table_field->field_index];
-      if (!hist->parse(mem_root, field, hist_type,
+      if (!hist->parse(mem_root, db_name->str, table_name->str,
+                       field, hist_type,
                        val.ptr(), val.length()))
       {
         table_field->read_stats->histogram= hist;
@@ -1247,9 +1248,9 @@ class Column_stat: public Stat_table
 };
 
 
-bool Histogram_binary::parse(MEM_ROOT *mem_root, Field*,
-                             Histogram_type type_arg, const char *hist_data,
-                             size_t hist_data_len)
+bool Histogram_binary::parse(MEM_ROOT *mem_root, const char*, const char*,
+                             Field*, Histogram_type type_arg,
+                             const char *hist_data, size_t hist_data_len)
 {
   /* On-disk an in-memory formats are the same. Just copy the data. */
   type= type_arg;

sql/sql_statistics.h

@@ -154,7 +154,9 @@ class Histogram_builder;
 class Histogram_base
 {
 public:
-  virtual bool parse(MEM_ROOT *mem_root, Field *field, Histogram_type type_arg,
+  virtual bool parse(MEM_ROOT *mem_root,
+                     const char *db_name, const char *table_name,
+                     Field *field, Histogram_type type_arg,
                      const char *hist_data, size_t hist_data_len)= 0;
   virtual void serialize(Field *to_field)= 0;
 
@@ -311,8 +313,9 @@ class Histogram_binary : public Histogram_base
 
   Histogram_type get_type() override { return type; }
 
-  bool parse(MEM_ROOT *mem_root, Field *, Histogram_type type_arg,
-             const char *hist_data, size_t hist_data_len) override;
+  bool parse(MEM_ROOT *mem_root, const char*, const char*, Field*,
+             Histogram_type type_arg, const char *hist_data,
+             size_t hist_data_len) override;
   void serialize(Field *to_field) override;
   void init_for_collection(MEM_ROOT *mem_root, Histogram_type htype_arg,
                            ulonglong size) override;