It seems like we can have more than 1000 columns. I create a testing survey. 16 groups. each group contains 1 multiple choice question. And each multiple question contains 100 items. So I have around 1600 columns in the result table. It seems fine. (i.e. I can test it and export the data).

Am I suppose to see error message when I reach the limit of column? If anyone have experience of reaching limit. Please share. I use local php server (easyPHP: APACHE 2.4.4 MYSQL 5.6.11 PHP 5.5.0)

You can always test if you reach the column limit by (temporarily) activating the survey.

I tested it as I mentioned previously. I think I know why I can have more than 1000 columns. It seems Limesurvey use MyISAM as default storage engine. (See database_helper.php: $sOptions='ENGINE=MyISAM DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci'

Just wondering why not use InnoDB as default as it has many more features? (i.e. Transactions, Row Locking, Relational base design , etc).

Just a thought

To my knowledge InnoDB uses more bytes per column and thus limits the amount of total questions even more which is the main reason we are still using MyISAM.

i think lots of problems will happen if we have no a high level mysql design.
we are in big data or die

InnoDB only supports 1000 columns at a maximum.

With MyISAM you have at least the chance to have more than 1000 - as already said it depends on the question types and number of subquestion how many you really have.
And, we don't need row locking or transactions - survey participants primarily write, administrators read. There is rarely a different scenario, so advanced InnoDB features are beaten by the faster MyISAM engine.

c_schmitz wrote: With MyISAM you have at least the chance to have more than 1000 - as alread said it depends on the question types and umber of subquestion how many you really have.

Near the max: 728 question + 670 sub question

in big data time, we always need a survey with over 1000 questions,i think myisam is good choice.
the problem is mysql database design, ls has no a robust database design foundation. eg, put all things in one table.that's not best database design concept.tha's almost the most important thing of a system.

"Big data" usually isn't collected in survey style but comes from other sources.

Just out of curiosity: Who fills a 1000 question survey? I am sure this is is a very special case.

holch wrote: Just out of curiosity: Who fills a 1000 question survey? I am sure this is is a very special case.

If you use many conditions and filters you don't need to fill out 1000 questions.

Wonder if the php session file for every user would be even bigger with a 1000 questions survey than with simple surveys, where every user got a 1 to 1,6 MB session file created, when starting the survey. That is a problem with session files on tmp partitions.

For me "Big data" starts if a database is 500GB and more.

Would be interesting to see some database designs of other applications with survey elements.

Maybe not everyone will fill 1000 questions, but still, this is well beyond most questionnaires that I have ever seen.

Even an insane 1 hour CAPI questionnaire that we had to run some time ago should get anywhere close to that. Loads of conditions and paths depending on specific vehicle models that the respondent had to own would hide many of the questions and it still took easily 1 hour to fill, even with experienced interviewers. I need to find this monster again, but my guess would be about 200-300 questions.

By the way: are you a sleep walker? Its 3am over there already... go and sleep!