mayfly-go/server/internal/db/dbm/postgres/dialect.go

package postgres

import (
	"database/sql"
	"fmt"
	"mayfly-go/internal/db/dbm/dbi"
	"mayfly-go/pkg/errorx"
	"mayfly-go/pkg/utils/anyx"
	"mayfly-go/pkg/utils/collx"
	"regexp"
	"strings"
	"time"
)

const (
	PGSQL_META_FILE      = "metasql/pgsql_meta.sql"
	PGSQL_DB_SCHEMAS     = "PGSQL_DB_SCHEMAS"
	PGSQL_TABLE_INFO_KEY = "PGSQL_TABLE_INFO"
	PGSQL_INDEX_INFO_KEY = "PGSQL_INDEX_INFO"
	PGSQL_COLUMN_MA_KEY  = "PGSQL_COLUMN_MA"
	PGSQL_TABLE_DDL_KEY  = "PGSQL_TABLE_DDL_FUNC"
)

type PgsqlDialect struct {
	dc *dbi.DbConn
}

func (pd *PgsqlDialect) GetDbServer() (*dbi.DbServer, error) {
	_, res, err := pd.dc.Query("SELECT version() as server_version")
	if err != nil {
		return nil, err
	}
	ds := &dbi.DbServer{
		Version: anyx.ConvString(res[0]["server_version"]),
	}
	return ds, nil
}

func (pd *PgsqlDialect) GetDbNames() ([]string, error) {
	_, res, err := pd.dc.Query("SELECT datname AS dbname FROM pg_database WHERE datistemplate = false AND has_database_privilege(datname, 'CONNECT')")
	if err != nil {
		return nil, err
	}

	databases := make([]string, 0)
	for _, re := range res {
		databases = append(databases, anyx.ConvString(re["dbname"]))
	}

	return databases, nil
}

// 获取表基础元信息, 如表名等
func (pd *PgsqlDialect) GetTables() ([]dbi.Table, error) {
	_, res, err := pd.dc.Query(dbi.GetLocalSql(PGSQL_META_FILE, PGSQL_TABLE_INFO_KEY))
	if err != nil {
		return nil, err
	}

	tables := make([]dbi.Table, 0)
	for _, re := range res {
		tables = append(tables, dbi.Table{
			TableName:    re["tableName"].(string),
			TableComment: anyx.ConvString(re["tableComment"]),
			CreateTime:   anyx.ConvString(re["createTime"]),
			TableRows:    anyx.ConvInt(re["tableRows"]),
			DataLength:   anyx.ConvInt64(re["dataLength"]),
			IndexLength:  anyx.ConvInt64(re["indexLength"]),
		})
	}
	return tables, nil
}

// 获取列元信息, 如列名等
func (pd *PgsqlDialect) GetColumns(tableNames ...string) ([]dbi.Column, error) {
	dbType := pd.dc.Info.Type
	tableName := strings.Join(collx.ArrayMap[string, string](tableNames, func(val string) string {
		return fmt.Sprintf("'%s'", dbType.RemoveQuote(val))
	}), ",")

	_, res, err := pd.dc.Query(fmt.Sprintf(dbi.GetLocalSql(PGSQL_META_FILE, PGSQL_COLUMN_MA_KEY), tableName))
	if err != nil {
		return nil, err
	}

	columns := make([]dbi.Column, 0)
	for _, re := range res {
		columns = append(columns, dbi.Column{
			TableName:     anyx.ConvString(re["tableName"]),
			ColumnName:    anyx.ConvString(re["columnName"]),
			ColumnType:    anyx.ConvString(re["columnType"]),
			ColumnComment: anyx.ConvString(re["columnComment"]),
			Nullable:      anyx.ConvString(re["nullable"]),
			IsPrimaryKey:  anyx.ConvInt(re["isPrimaryKey"]) == 1,
			IsIdentity:    anyx.ConvInt(re["isIdentity"]) == 1,
			ColumnDefault: anyx.ConvString(re["columnDefault"]),
			NumScale:      anyx.ConvString(re["numScale"]),
		})
	}
	return columns, nil
}

func (pd *PgsqlDialect) GetPrimaryKey(tablename string) (string, error) {
	columns, err := pd.GetColumns(tablename)
	if err != nil {
		return "", err
	}
	if len(columns) == 0 {
		return "", errorx.NewBiz("[%s] 表不存在", tablename)
	}
	for _, v := range columns {
		if v.IsPrimaryKey {
			return v.ColumnName, nil
		}
	}

	return columns[0].ColumnName, nil
}

// 获取表索引信息
func (pd *PgsqlDialect) GetTableIndex(tableName string) ([]dbi.Index, error) {
	_, res, err := pd.dc.Query(fmt.Sprintf(dbi.GetLocalSql(PGSQL_META_FILE, PGSQL_INDEX_INFO_KEY), tableName))
	if err != nil {
		return nil, err
	}

	indexs := make([]dbi.Index, 0)
	for _, re := range res {
		indexs = append(indexs, dbi.Index{
			IndexName:    anyx.ConvString(re["indexName"]),
			ColumnName:   anyx.ConvString(re["columnName"]),
			IndexType:    anyx.ConvString(re["IndexType"]),
			IndexComment: anyx.ConvString(re["indexComment"]),
			IsUnique:     anyx.ConvInt(re["isUnique"]) == 1,
			SeqInIndex:   anyx.ConvInt(re["seqInIndex"]),
		})
	}
	// 把查询结果以索引名分组，索引字段以逗号连接
	result := make([]dbi.Index, 0)
	key := ""
	for _, v := range indexs {
		// 当前的索引名
		in := v.IndexName
		if key == in {
			// 索引字段已根据名称和顺序排序，故取最后一个即可
			i := len(result) - 1
			// 同索引字段以逗号连接
			result[i].ColumnName = result[i].ColumnName + "," + v.ColumnName
		} else {
			key = in
			result = append(result, v)
		}
	}
	return result, nil
}

// 获取建表ddl
func (pd *PgsqlDialect) GetTableDDL(tableName string) (string, error) {
	_, err := pd.dc.Exec(dbi.GetLocalSql(PGSQL_META_FILE, PGSQL_TABLE_DDL_KEY))
	if err != nil {
		return "", err
	}

	ddlSql := fmt.Sprintf("select showcreatetable('%s','%s') as sql", pd.currentSchema(), tableName)
	_, res, err := pd.dc.Query(ddlSql)
	if err != nil {
		return "", err
	}

	return res[0]["sql"].(string), nil
}

// 获取pgsql当前连接的库可访问的schemaNames
func (pd *PgsqlDialect) GetSchemas() ([]string, error) {
	sql := dbi.GetLocalSql(PGSQL_META_FILE, PGSQL_DB_SCHEMAS)
	_, res, err := pd.dc.Query(sql)
	if err != nil {
		return nil, err
	}
	schemaNames := make([]string, 0)
	for _, re := range res {
		schemaNames = append(schemaNames, anyx.ConvString(re["schemaName"]))
	}
	return schemaNames, nil
}

// GetDbProgram 获取数据库程序模块，用于数据库备份与恢复
func (pd *PgsqlDialect) GetDbProgram() (dbi.DbProgram, error) {
	return nil, fmt.Errorf("该数据库类型不支持数据库备份与恢复: %v", pd.dc.Info.Type)
}

func (pd *PgsqlDialect) BatchInsert(tx *sql.Tx, tableName string, columns []string, values [][]any, duplicateStrategy int) (int64, error) {
	// 执行批量insert sql，跟mysql一样  pg或高斯支持批量insert语法
	// insert into table_name (column1, column2, ...) values (value1, value2, ...), (value1, value2, ...), ...

	// 把二维数组转为一维数组
	var args []any
	for _, v := range values {
		args = append(args, v...)
	}

	// 构建占位符字符串 "($1, $2, $3), ($4, $5, $6), ..." 用于指定参数
	var placeholders []string
	for i := 0; i < len(args); i += len(columns) {
		var placeholder []string
		for j := 0; j < len(columns); j++ {
			placeholder = append(placeholder, fmt.Sprintf("$%d", i+j+1))
		}
		placeholders = append(placeholders, "("+strings.Join(placeholder, ", ")+")")
	}

	// 根据冲突策略生成后缀
	suffix := ""
	if pd.dc.Info.Type == dbi.DbTypeGauss {
		// 高斯db使用ON DUPLICATE KEY UPDATE 语法参考 https://support.huaweicloud.com/distributed-devg-v3-gaussdb/gaussdb-12-0607.html#ZH-CN_TOPIC_0000001633948138
		suffix = pd.gaussOnDuplicateStrategySql(duplicateStrategy, tableName, columns)
	} else {
		// pgsql 默认使用 on conflict 语法参考 http://www.postgres.cn/docs/12/sql-insert.html
		// vastbase语法参考 https://docs.vastdata.com.cn/zh/docs/VastbaseE100Ver3.0.0/doc/SQL%E8%AF%AD%E6%B3%95/INSERT.html
		// kingbase语法参考 https://help.kingbase.com.cn/v8/development/sql-plsql/sql/SQL_Statements_9.html#insert
		suffix = pd.pgsqlOnDuplicateStrategySql(duplicateStrategy, tableName, columns)
	}

	sqlStr := fmt.Sprintf("insert into %s (%s) values %s %s", pd.dc.Info.Type.QuoteIdentifier(tableName), strings.Join(columns, ","), strings.Join(placeholders, ", "), suffix)
	// 执行批量insert sql

	return pd.dc.TxExec(tx, sqlStr, args...)
}

// pgsql默认唯一键冲突策略
func (pd PgsqlDialect) pgsqlOnDuplicateStrategySql(duplicateStrategy int, tableName string, columns []string) string {
	suffix := ""
	if duplicateStrategy == dbi.DuplicateStrategyIgnore {
		suffix = " \n on conflict do nothing"
	} else if duplicateStrategy == dbi.DuplicateStrategyUpdate {
		// 生成 on conflict () do update set column1 = excluded.column1, column2 = excluded.column2, ...
		var updateColumns []string
		for _, col := range columns {
			updateColumns = append(updateColumns, fmt.Sprintf("%s = excluded.%s", col, col))
		}
		// 查询唯一键名,拼接冲突sql
		_, keyRes, _ := pd.dc.Query("SELECT constraint_name FROM information_schema.table_constraints WHERE constraint_schema = $1 AND table_name = $2 AND constraint_type in ('PRIMARY KEY', 'UNIQUE') ", pd.currentSchema(), tableName)
		if len(keyRes) > 0 {
			for _, re := range keyRes {
				key := anyx.ToString(re["constraint_name"])
				if key != "" {
					suffix += fmt.Sprintf(" \n on conflict on constraint %s do update set %s \n", key, strings.Join(updateColumns, ", "))
				}
			}
		}
	}
	return suffix
}

// 高斯db唯一键冲突策略,使用ON DUPLICATE KEY UPDATE 参考：https://support.huaweicloud.com/distributed-devg-v3-gaussdb/gaussdb-12-0607.html#ZH-CN_TOPIC_0000001633948138
func (pd PgsqlDialect) gaussOnDuplicateStrategySql(duplicateStrategy int, tableName string, columns []string) string {
	suffix := ""
	if duplicateStrategy == dbi.DuplicateStrategyIgnore {
		suffix = " \n ON DUPLICATE KEY UPDATE NOTHING"
	} else if duplicateStrategy == dbi.DuplicateStrategyUpdate {

		// 查出表里的唯一键涉及的字段
		var uniqueColumns []string
		indexs, err := pd.GetTableIndex(tableName)
		if err == nil {
			for _, index := range indexs {
				if index.IsUnique {
					cols := strings.Split(index.ColumnName, ",")
					for _, col := range cols {
						if !collx.ArrayContains(uniqueColumns, strings.ToLower(col)) {
							uniqueColumns = append(uniqueColumns, strings.ToLower(col))
						}
					}
				}
			}
		}

		suffix = " \n ON DUPLICATE KEY UPDATE "
		for i, col := range columns {
			// ON DUPLICATE KEY UPDATE语句不支持更新唯一键字段，所以得去掉
			if !collx.ArrayContains(uniqueColumns, pd.dc.Info.Type.RemoveQuote(strings.ToLower(col))) {
				suffix += fmt.Sprintf("%s = excluded.%s", col, col)
				if i < len(columns)-1 {
					suffix += ", "
				}
			}
		}
	}
	return suffix
}

// 从连接信息中获取数据库和schema信息
func (pd *PgsqlDialect) currentSchema() string {
	dbName := pd.dc.Info.Database
	schema := ""
	arr := strings.Split(dbName, "/")
	if len(arr) == 2 {
		schema = arr[1]
	}
	return schema
}

func (pd *PgsqlDialect) GetDataConverter() dbi.DataConverter {
	return converter
}

var (
	// 数字类型
	numberRegexp = regexp.MustCompile(`(?i)int|double|float|number|decimal|byte|bit`)
	// 日期时间类型
	datetimeRegexp = regexp.MustCompile(`(?i)datetime|timestamp`)
	// 日期类型
	dateRegexp = regexp.MustCompile(`(?i)date`)
	// 时间类型
	timeRegexp = regexp.MustCompile(`(?i)time`)

	converter = new(DataConverter)
)

type DataConverter struct {
}

func (dc *DataConverter) GetDataType(dbColumnType string) dbi.DataType {
	if numberRegexp.MatchString(dbColumnType) {
		return dbi.DataTypeNumber
	}
	// 日期时间类型
	if datetimeRegexp.MatchString(dbColumnType) {
		return dbi.DataTypeDateTime
	}
	// 日期类型
	if dateRegexp.MatchString(dbColumnType) {
		return dbi.DataTypeDate
	}
	// 时间类型
	if timeRegexp.MatchString(dbColumnType) {
		return dbi.DataTypeTime
	}
	return dbi.DataTypeString
}

func (dc *DataConverter) FormatData(dbColumnValue any, dataType dbi.DataType) string {
	str := fmt.Sprintf("%v", dbColumnValue)
	switch dataType {
	case dbi.DataTypeDateTime: // "2024-01-02T22:16:28.545377+08:00"
		res, _ := time.Parse(time.RFC3339, str)
		return res.Format(time.DateTime)
	case dbi.DataTypeDate: //  "2024-01-02T00:00:00Z"
		res, _ := time.Parse(time.RFC3339, str)
		return res.Format(time.DateOnly)
	case dbi.DataTypeTime: // "0000-01-01T22:16:28.545075+08:00"
		res, _ := time.Parse(time.RFC3339, str)
		return res.Format(time.TimeOnly)
	}
	return anyx.ConvString(dbColumnValue)
}

func (dc *DataConverter) ParseData(dbColumnValue any, dataType dbi.DataType) any {
	// 如果dataType是datetime而dbColumnValue是string类型，则需要转换为time.Time类型
	_, ok := dbColumnValue.(string)
	if dataType == dbi.DataTypeDateTime && ok {
		res, _ := time.Parse(time.RFC3339, anyx.ToString(dbColumnValue))
		return res
	}
	if dataType == dbi.DataTypeDate && ok {
		res, _ := time.Parse(time.DateOnly, anyx.ToString(dbColumnValue))
		return res
	}
	if dataType == dbi.DataTypeTime && ok {
		res, _ := time.Parse(time.TimeOnly, anyx.ToString(dbColumnValue))
		return res
	}
	return dbColumnValue
}

func (pd *PgsqlDialect) IsGauss() bool {
	return strings.Contains(pd.dc.Info.Params, "gauss")
}

func (pd *PgsqlDialect) CopyTable(copy *dbi.DbCopyTable) error {
	tableName := copy.TableName
	// 生成新表名,为老表明+_copy_时间戳
	newTableName := tableName + "_copy_" + time.Now().Format("20060102150405")
	// 执行根据旧表创建新表
	_, err := pd.dc.Exec(fmt.Sprintf("create table %s (like %s)", newTableName, tableName))
	if err != nil {
		return err
	}

	// 复制数据
	if copy.CopyData {
		go func() {
			_, _ = pd.dc.Exec(fmt.Sprintf("insert into %s select * from %s", newTableName, tableName))
		}()
	}

	// 查询旧表的自增字段名 重新设置新表的序列序列器
	_, res, err := pd.dc.Query(fmt.Sprintf("select column_name from information_schema.columns where table_name = '%s' and column_default like 'nextval%%'", tableName))
	if err != nil {
		return err
	}

	for _, re := range res {
		colName := anyx.ConvString(re["column_name"])
		if colName != "" {

			// 查询自增列当前最大值
			_, maxRes, err := pd.dc.Query(fmt.Sprintf("select max(%s) max_val from %s", colName, tableName))
			if err != nil {
				return err
			}
			maxVal := anyx.ConvInt(maxRes[0]["max_val"])
			// 序列起始值为1或当前最大值+1
			if maxVal <= 0 {
				maxVal = 1
			} else {
				maxVal += 1
			}

			// 之所以不用tableName_colName_seq是因为gauss会自动创建同名的序列，且无法修改序列起始值，所以直接使用新序列值
			newSeqName := fmt.Sprintf("%s_%s_copy_seq", newTableName, colName)

			// 创建自增序列，当前最大值为旧表最大值
			_, err = pd.dc.Exec(fmt.Sprintf("CREATE SEQUENCE %s START %d INCREMENT 1", newSeqName, maxVal))
			if err != nil {
				return err
			}
			// 将新表的自增主键序列与主键列相关联
			_, err = pd.dc.Exec(fmt.Sprintf("alter table %s alter column %s set default nextval('%s')", newTableName, colName, newSeqName))
			if err != nil {
				return err
			}
		}
	}
	return err
}