Comparison of database column types in MySQL, PostgreSQL, and SQLite? (Cross-Mapping)

Question

I am trying to find some way to relate column types across the the most used Databases: MySQL, PostgreSQL, and SQLite.

Here is what I have so far, but I'm afraid it's not done and I need some people with more experience to help me finish any missing types.

MySQL                   PostgreSQL          SQLite

TINYINT                 SMALLINT            INTEGER
SMALLINT                SMALLINT
MEDIUMINT               INTEGER
BIGINT                  BIGINT
BIT                     BIT                 INTEGER
_______________________________________________________

TINYINT UNSIGNED        SMALLINT            INTEGER
SMALLINT UNSIGNED       INTEGER
MEDIUMINT UNSIGNED      INTEGER
INT UNSIGNED            BIGINT
BIGINT UNSIGNED         NUMERIC(20)
_______________________________________________________

DOUBLE                  DOUBLE PRECISION    REAL
FLOAT                   REAL                REAL
DECIMAL                 DECIMAL             REAL
NUMERIC                 NUMERIC             REAL
_______________________________________________________

BOOLEAN                 BOOLEAN             INTEGER
_______________________________________________________

DATE                    DATE                TEXT
TIME                    TIME
DATETIME                TIMESTAMP
_______________________________________________________

TIMESTAMP DEFAULT       TIMESTAMP DEFAULT   TEXT
NOW()                   NOW()   
_______________________________________________________

LONGTEXT                TEXT                TEXT
MEDIUMTEXT              TEXT                TEXT
BLOB                    BYTEA               BLOB
VARCHAR                 VARCHAR             TEXT
CHAR                    CHAR                TEXT
_______________________________________________________

columnname INT          columnname SERIAL   INTEGER PRIMARY 
AUTO_INCREMENT                              KEY AUTOINCREMENT

Solution

List of things I'd do differently:

MEDIUMINT in MySQL is an odd duck (3 bytes). I would avoid it, but otherwise map it to INTEGER too.

The MySQL BOOLEAN (alias BOOL, alias TINYINT(1) ) is not compatible with the pg boolean type. You may or may not be able to port apps depending on what they use as boolean literals. In MySQL, TRUE and FALSE map to 1 and 0 integer values. It looks like the pg BOOLEAN type uses string literal notation. So apps may or may not be portable - at least it is no drop in replacement.

Finally, for the last line in your tabl I think the SQLite phrase should read:

INTEGER PRIMARY KEY AUTOINCREMENT

This is roughly equivalent to

BIGINT PRIMARY KEY AUTO_INCREMENT

in MySQL. In postgres, the SERIAL datatype results in an INTEGER column, and this will about the same as MySQL's

INTEGER PRIMARY KEY AUTO_INCREMENT

Postgres also has a BIGSERIAL type, which is the same as SERIAL but with a BIGINT type instead of an INT type.

What I missed:

I am missing INTEGER (alias INT) for MySQL. It is comparable to INTEGER in pg. Very important omissions: VARCHAR and CHAR. Semantically, VARCHAR in MySQL and PG, and CHAR in MySQL and PG are the same, but in MySQL these types have a much shorter maximum length. In MySQL these types can have a maximum of a little less than 64kb, in pg 1Gb (bytes). The actual length specifier is expressed in the number of characters, so if you have a multi-byte character set, you have to divide the maximum lenght by the maximum number of characters to get the theoretical maximum length specified for that characterset. In SQLite, VARCHAR and CHAR map both to TEXT

The BIT datatypes in MySQL and PG have roughly the same semantics, but in MySQL the maximum length of the BIT data type is 64 (bits)

I think the MySQL VARBINARY data type is best comparable to PG's BYTEA datatype. (but indeed MySQL's BLOB types also map to that)

The FLOAT type in MySQL should be equivalent to REAL in postgres (and REAL in SQLite too) The DECIMAL type in MySQL is equivalent to DECIMAL in postgres, except that in postgres, the type does not impose an arbtrary limit on the precision, whereas in MySQL the maximum precision is (i believe) 70. (that is, 70 number positions) For both MySQL and Postgres, NUMERIC is an alias for the DECIMAL type.

Comments

[Evan Carroll]

there is another difference, in Pg you only use varchar() when you have a reasonable reason to invoke the constraint it provides. In MySQL you do it to have a large internally inlined blob of text that runs quickly. In Pg it runs slower, and takes up my room. In Pg you would almost never use varchar().

[Roland Bouman]

EvanCarrol, that's interesting. So what should one use in pg to store fairly small bits of text like person names, product names, short (say, less than 255) descriptions? just TEXT?

[user340140]

postgres booleans are not string literal types, they only look that way. if you cast them to integer the come out o,1, or NULL as apropriate, conversely if you have 0,1 or NULL integers you can cast them to boolean. COPY...FROM accepts numbers but INSERT needs an explicit cast or quotes around the number. so '0', cast ( 0 as boolen),'f', and false will all work in an insert expecting boolean.