SQL: Application cross-db maintain only one generic schema - mysql

Question

I actually have an application based on MySQL with a schema based on InnoDB (with constraints...)
My co-workers need to import this schema, so I export my schema in SQL files.
For example:
DROP TABLE IF EXISTS `admins`;
CREATE TABLE `admins` (
`id` smallint(5) unsigned NOT NULL AUTO_INCREMENT,
`username` varchar(45) NOT NULL,
`password` varchar(45) NOT NULL,
`email` varchar(45) DEFAULT NULL,
`creation_date` datetime NOT NULL,
`close_date` datetime DEFAULT NULL,
`close_reason` varchar(45) DEFAULT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 AUTO_INCREMENT=4 ;
Now, I would like to have a cross-db application, so:
I tried to import my previous sql files in PostgreSQL, but it didn't work, my SQL files are mysql-related (for example use of ` character...)
I tried to export my schema with mysqldump and a compatibility mode "--compatible=ansi" my goal: have a generic sql file compatible with all major SGBD. But it didn't work: PostgreSQL returns error about synthax
compatible=ansi returns:
DROP TABLE IF EXISTS "admins";
/*!40101 SET #saved_cs_client = ##character_set_client */;
/*!40101 SET character_set_client = utf8 */;
CREATE TABLE "admins" (
"id" smallint(5) unsigned NOT NULL AUTO_INCREMENT,
"username" varchar(45) NOT NULL,
"password" varchar(45) NOT NULL,
"email" varchar(45) DEFAULT NULL,
"creation_date" datetime NOT NULL,
"close_date" datetime DEFAULT NULL,
"close_reason" varchar(45) DEFAULT NULL,
PRIMARY KEY ("id")
);
/*!40101 SET character_set_client = #saved_cs_client */;
I even tried to export with compatibility=postgresql:
DROP TABLE IF EXISTS "admins";
/*!40101 SET #saved_cs_client = ##character_set_client */;
/*!40101 SET character_set_client = utf8 */;
CREATE TABLE "admins" (
"id" smallint(5) unsigned NOT NULL,
"username" varchar(45) NOT NULL,
"password" varchar(45) NOT NULL,
"email" varchar(45) DEFAULT NULL,
"creation_date" datetime NOT NULL,
"close_date" datetime DEFAULT NULL,
"close_reason" varchar(45) DEFAULT NULL,
PRIMARY KEY ("id")
);
/*!40101 SET character_set_client = #saved_cs_client */;
But also didn't work...
I know there are tools to convert MySQL schema to PostgreSQL schema but this isn't the goal...
My question: Is it possible to have only one SQL file compatible with MySQL, PostgreSQL, SQLite... and don't maintain a SQL file for each SGBD ?
Thank you

Answer 1

My question: Is it possible to have only one SQL file compatible with MySQL, PostgreSQL, SQLite... and don't maintain a SQL file for each SGBD ?
Not easily with raw SQL, unless you wish to use a pathetic subset of the databases' supported features.
SELECTs and DML in SQL can be moderately portable, but DDL is generally a hopeless nightmare for all but the total basics. You'll want an abstraction tool that generates the SQL for you, handling database specific differences in sequences/generated keys, type naming, constraints, index creation, etc.
As just one example, lets look at auto-incrementing values / sequences, as frequently used for synthetic keys:
MySQL: integer AUTO_INCREMENT
PostgreSQL: SERIAL (shorthand for a sequence)
MS-SQL: int IDENTITY(1,1)
Oracle (below 12c): No direct support, use a sequence.
Oracle (12c and above): NUMBER GENERATED BY DEFAULT ON NULL AS IDENTITY
.. and that's just for the very common task of a generated key. Lots of other fun differences exist. For example, MySQL has tinyint and unsigned int. PostgreSQL does not. PostgreSQL has bool and has bit(n) bitfields, range-types, PostGIS types, etc etc etc which most other DBs don't have. Even for things that're shared, quirks abound - specifying "4 byte signed integer" across all DBs isn't even trivial.
One option to help is Liquibase which I've heard good things about. Some people instead use an ORM to manage their DDL generation instead - though those tend to use, again, only the most primitive of database features.