fix empty form elements

There is no way to fix it in form_checkboxes() itself afaik, because even without htmlspecialchars, it would still come out as "Array". You can only create arrays indirectly through form submissions, by naming your values "foo[]". But you can not create empty arrays this way.

So it is really up to the module which uses form_checkboxes() to handle this. A simple is_array() should do the trick.

system_settings_save can not really cope with this situation. This makes a problem with empty node options. And the fact that we have a check everywhere we use form_checkboxes is error prone, too. Here is a patch.

If this accepted I'll comb the core for dead code. For eg. the first three lines of taxonomy_save_vocabulary and so.

Oh NO! One strayed space! I fixed it before steven beheads me....

Among other changes, I have moved the hidden field form_checkboxes[] out of $edit. I have included HTML code, yes, but this is a very special thing, so it does not merit changing form_input just 'cos of this. And you won't need to theme a hidden field...

killes did not like the previous version.

Dries?

I have to agree with Steven that it is up to the calling code...and I hardly think this is critical. Code that is expecting an array from form_checkboxes() should ensure that it is dealing with an array (i.e. is_array()). As for dealing with variables set through things like system_settings_save(), the same thing applies. The code that is retrieving the variable should ensure that it is working with an array.

With regards to your patch, what's the difference if the calling code has to remember to check via is_array() or remember to call fix_checkboxes()? This method is no less error prone as it's just as easy to forget this call as it is to check for an array. That is unless you are suggesting that fix_checkboxes() gets called for every page load which seems rather crufty.

I wrote the original patch to add form_checkboxes() -- which didn't exist -- and if I remember, this was the desired behavior if nothing was selected. Isn't this the same behavior one would see using a multiple form_select() if the user chooses nothing? They should behave the same. A single select widget is the same as a radio group, and a multiple select widget is the same as a group of checkboxes. They should behave the same; the select types just take up less realestate.

Previous patch had index.php missing.

jhriggs, the current problem is that a form_checkboxes expected to be an array. Currently, empty form_checkboxes is a string. That's bad.

-1 on this patch.

I still don't understand the problem. What do you mean by "form_checkboxes expected to be an array"? Is there code in core that assumes it is an array but it is not checked with is_array()? If so, let's add the missing checks. The same goes for contrib. This seems like swatting a fly with a sledge hammer to me.

If we do go ahead and make this change, however, then we will need to change form_select() also (when multiple = TRUE). Radios and single select should be consistent, as should checkboxes and multiple select.

See Gordon's complaint http://lists.drupal.org/archives/drupal-devel/2005-04/msg00934.html

Forgot to change the title.

It seems that it's not obvious why this is necessary. The calling code can not always handle itself the situation if the save is done by system_save_settings. The edge case is: I set up something, so variable_get no longer falls to its default value, later I decide to empty it. If we do not do something like this patch, system_save_settings can not know there is something to be set to 0/array() because it is just not in $_POST.

Small bugfix.

doh. still has serious problems when the name is an array. stay tuned.

Recursive version.

Patch looks good but I would add some phpdoc to document what and why we are doing this.

People told me that form_radios does not need this because we can reasonably expect that a radio is checked, that's different between a bunch of checkboxes and a bunch of radios.

form_select submits an empty array (as also told by some).

So we are back to checkbox and checkboxes which need the fix.

Committed to HEAD. Thanks.

Why do we now have both a _fix_checkboxes() and fix_checkboxes() function? Shouldn't we name these something a little more intuitive? Without poking through the code, I have no idea what _fix_checkboxes is for and why it's different than fix_checkboxes. It really deserves some phpdoc too. Any sort of more-pleasant variable naming in _fix_checkboxes() would be much appreciated as well... It's easy to get lost in that array1/array2/single letter variable business pretty fast. Sorry for nit-picking, but most of the Drupal core code is far more readable than this.

It's strange that this problem still appears in drupal 4.6.3 (August 15, 2005); 4.6 CVS, and CVS Head and has been marked fixed on May 21, 2005. The last patch (fix_form_3.patch from comment 20) still applies with little fuzziness and fixes the problem.

It's strange that this problem still appears in drupal 4.6.3 (August 15, 2005); 4.6 CVS, and CVS Head and has been marked fixed on May 21, 2005. The last patch (fix_form_3.patch from comment 20) still applies with little fuzziness and fixes the problem.

The problem has been fixed in HEAD as Dries said in #21 HEAD refer to the future 4.7 version.

Here is the updated patch that apply to 4.6.3

Can someone test this first?

For me it works on a site with 2000 - 3000 visitors/day. Maybe it's a good idea to also add this sql statement in updates.inc?
update variable set value="a:0:{}" where name like "node_options_%" and value = "s:1:\"0\";";

Setting to won't fix, unless someone tests the patch.

1) doesn't seem to have an adverse effect.
2) in a way is a nit picky issue.
3) altering the behavior of form_checkboxes 5 revisions into the release may cause more harm than good.
4) is_array is probably far less costly performance wise than this patch.

feel free to change the status if I am incorrect. I'm just cleaning out old issues.

oh yeah, actually setting to won't fix.

Ok, I don't know how to do this patch thing, but couldn't this be fixed simply by adding the following:

settype($node_options, "array");

In my node.module file, it's line 1314. It's directly after "$node_options = ..." in "function node_form(...)"

I'm going to set this to active, because I think this will fix the problem. Maybe it's fixed in 4.7, but many of us are still stuck in 4.6, and I'd like to get this dealt with.

There's also a problem in the "node_validate($node)" function:

It's line 1248 in my node.module.

  else {
    // Validate for normal users:
    $node->uid = $user->uid ? $user->uid : 0;
    // Force defaults in case people modify the form:
    /* the following line is added */
    settype($node_options, "array");
    $node_options = variable_get('node_options_'. $node->type, array('status', 'promote'));
    $node->status = in_array('status', $node_options);
    $node->moderate = in_array('moderate', $node_options);
    $node->promote = in_array('promote', $node_options);
    $node->sticky = in_array('sticky', $node_options);
    $node->revision = in_array('revision', $node_options);
    unset($node->created);

This has been fixed in 4.7 HEAD.

-K

#30 works for me. I am attaching the corresponding patch, tested in 4.6.6.
This is not the general solution, but fixes this related bug.

This is fixed in 4.7 but might still be an issue for 4.6 users.