[GRASS-dev] r54684 - db_sqltype_name(): report variable-width string instead of fixed-width n-character string

[Markus Metz]

On Mon, Mar 25, 2013 at 2:09 PM, Vaclav Petras <wenzeslaus at gmail.com> wrote:
> Hi,
>
> On 25 March 2013 09:49, Markus Metz <markus.metz.giswork at gmail.com> wrote:
>> On Sat, Mar 23, 2013 at 5:37 PM, Martin Landa <landa.martin at gmail.com> wrote:
>>> Hi,
>>>
>>> 2013/3/21 Markus Metz <markus.metz.giswork at gmail.com>:
>>>> what is the reason for r54684? It breaks a core functionality, i.e.
>>>> the output of vector_columns in lib/python/scripts/vector.py has now
>>>> changed and various scripts rely on that output. I have reverted
>>>> r54684 in 55475. The commit comment does not make sense to me because
>>>> db_sqltype_name() has never returned the width or precision.
>>>
>>> this change was related to `v.out.postgis` development. Hm, you are
>>> right. I was not aware enough of possible consequences. I agree with
>>> this revert.
>>
>> I noticed that 'CHARACTER' is used whenever the field width is known,
>> otherwise 'TEXT' is used. IOW, TEXT is what you probably mean with
>> CHARACTER VARYING.
>
>> Note that e.g. within sqlite, varchar becomes TEXT
>> and varchar (%d) becomes CHARACTER.
>>
> With sqlite3, all CHARACTER(%d)s, VARCHAR(%d)s, etc. become TEXT. [1]

This is the sqlite affinity, but the GRASS sqlite driver checks first
the declared datatype with sqlite3_column_decltype(), then the
affinity type if decltype is not available. Sqlite type and
DB_SQL_TYPE_* are not identical.

Markus M

>
> Vaclav
>
> [1] http://www.sqlite.org/datatype3.html