Jay,
> The question is, can we put this into a SQL database just as a
> flat file, ie as one table, with one column per question, or should
> we break it into several tables, for instance with a table of
> the "keywords" (the boxes that can be checked). Is there any
> problem on a typical Linux box with running PostgreSql database
> with one table of 120 columns?
Well, administration with that kind of a structure will be rather a pain. To
answer your question in 2 parts:
1) Yes, you can load the flat file with unchanged structure.
2) Yes, you will soon wish to use something better, such as:
(incomplete pseudo-SQL)
table survey_responses (
user_id
response_date
ip_address
status
key (user_id, response_date)
)
table survey_questions (
question_id
question_type
question_text
)
table survey_answers (
user_id
response_date
question_id
response
key (user_id, response_date, question_id)
)
This will make *much* more efficient use of DB resources, but requires you to
understand relational database design.
Incidentally, among my (many) unfinished projects is to publish the structure
and code backing the OpenOffice.org User Survey as an OSS survey-generation
engine. Someday ...
--
-Josh Berkus
Aglio Database Solutions
San Francisco
