Re: [Modeling-users] varchar and text

[Sebastien B. <sbi...@us...>]

Ernesto Revilla <er...@si...> wrote:
> Going thru the PostgreSQL doc version 7.3.2 it says:
> * varchar w/o upper limit is not standard SQL
> * text is postgres specific
> * varchar w/o limit and text have no performance penalities compared with
> varchar with limit
> * a lot of modern dbms systems support unlimited text.
>=20
> What's about mySQL? I would like to see if the the width requirement of
> varchar would be loosened (perhaps a warning, or ignored). I don't know h=
ow
> other database system treat this.

MySQL: http://www.mysql.com/doc/en/CHAR.html, no CHAR/VARCHAR w/o width
SQLite: accepts varchar without limit

  ModelValidation issues an error when a varchar column is declared
  without a width, however that does not prevent you to
  mdl_generate_DB_schema: if the db underneath supports it, everything
  will be fine :)

I agree, however, that ModelValidation should reflect the db
specificities when possible. This will be changed in a near future so
that it issues a message at the 'info' level (rather than 'error') for
rdbms supporting it.

  See also the note I recently added to the User's Guide about the
  'width' parameter (will be online for the next release), depicting the
  role played by 'width' at runtime:

<< Note: The field width is used at runtime when validating a string
   attribute's value (see 3.4, ``Validation''), before objects are saved
   in the database. SQL data types CHAR and VARCHAR require that width
   is specified; on the other hand, TEXT -when supported by the
   database- does not accept that width is set. If you set it on an
   attribute, it will be ignored when the database schema is generated
   (i.e. a TEXT field my_text with width=3D30 won't be declared as
   TEXT(30), but simply as TEXT), but it will be checked at runtime and
   validation for this attribute will fail if its value's length exceeds
   the given width. Note that the same goal can be achieved by writing a
   specific validation method for the attribute (see 3.4.2).
>>

-- S=E9bastien.