September 2005 - Development - lists.drupal.org

[drupal-devel] stepping through the revision history with AJAX
by Moshe Weitzman 08 Sep '05

08 Sep '05

there are 2 fantastic greasemonkey scripts out there which use AJAX to view the revision history of a wikipedia page. It would be great if we offerred this as a Contrib module, or even in core. Our AJAX and revision APIs have all the plumbing already. If you are busy, just watch this Flash movie - . Otherwise, see http://www.waxy.org/archive/2005/06/27/wikipedi.shtml If you are new to Greasemonkey, see http://www.google.com/url?sa=t&ct=res&cd=1&url=http%3A//greasemonkey.mozdev…

2 1

[drupal-devel] [bug] fix empty form elements
by tostinni 08 Sep '05

08 Sep '05

Issue status update for http://drupal.org/node/18663 Post a follow up: http://drupal.org/project/comments/add/18663 Project: Drupal -Version: 4.6.0 +Version: 4.6.3 Component: base system Category: bug reports Priority: critical Assigned to: chx Reported by: killes(a)www.drop.org Updated by: tostinni -Status: active +Status: patch (ready to be committed) Attachment: http://drupal.org/files/issues/fix_form-4.6.3.patch (3.54 KB) "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 tostinni Previous comments: ------------------------------------------------------------------------ Thu, 10 Mar 2005 01:51:56 +0000 : killes(a)www.drop.org If you select nothing from a set of checkboxes an empty string is save to the variables table. This is a problem if the calling code assumes, it would get an array. We should make it save an empty array. The problem is the form_hidden field that we add. even if we replace form_hidden($name, 0) by form_hidden($name, array()) it isn't going to work, because the empty array is eaten by htmlspecialchars. redLED found this bug and he does not like it: 02:53 < redLED> i just like things saying 'CRITICAL' in red 02:53 < redLED> it gives this warm impending doom feeling 02:53 < redLED> and is generally associated with nuclear reactors in people's subconsciousness Suggestions welcome. ------------------------------------------------------------------------ Sat, 12 Mar 2005 18:47:42 +0000 : Steven 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. ------------------------------------------------------------------------ Mon, 28 Mar 2005 16:27:03 +0000 : chx Attachment: http://drupal.org/files/issues/checkboxes.patch (1.71 KB) 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. ------------------------------------------------------------------------ Mon, 28 Mar 2005 16:32:16 +0000 : chx Attachment: http://drupal.org/files/issues/checkboxes_0.patch (1.71 KB) Oh NO! One strayed space! I fixed it before steven beheads me.... ------------------------------------------------------------------------ Mon, 28 Mar 2005 17:36:01 +0000 : chx Attachment: http://drupal.org/files/issues/checkboxes_1.patch (1.64 KB) 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... ------------------------------------------------------------------------ Fri, 08 Apr 2005 07:19:28 +0000 : chx Attachment: http://drupal.org/files/issues/form_checkboxes_2.patch (1.9 KB) killes did not like the previous version. ------------------------------------------------------------------------ Mon, 11 Apr 2005 23:15:05 +0000 : chx Dries? ------------------------------------------------------------------------ Tue, 12 Apr 2005 19:15:52 +0000 : chx Attachment: http://drupal.org/files/issues/form_checkboxes_3.patch (1.64 KB) ------------------------------------------------------------------------ Tue, 12 Apr 2005 19:39:00 +0000 : jhriggs 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. ------------------------------------------------------------------------ Tue, 12 Apr 2005 21:23:06 +0000 : chx Attachment: http://drupal.org/files/issues/form_checkboxes_4.patch (2.06 KB) 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. ------------------------------------------------------------------------ Tue, 12 Apr 2005 22:51:25 +0000 : jhriggs -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. ------------------------------------------------------------------------ Sat, 23 Apr 2005 11:35:30 +0000 : chx Attachment: http://drupal.org/files/issues/fix_form.patch (4.71 KB) See Gordon's complaint http://lists.drupal.org/archives/drupal-devel/2005-04/msg00934.html ------------------------------------------------------------------------ Sat, 23 Apr 2005 11:38:07 +0000 : chx Forgot to change the title. ------------------------------------------------------------------------ Sat, 23 Apr 2005 11:46:46 +0000 : chx 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. ------------------------------------------------------------------------ Sat, 23 Apr 2005 13:23:37 +0000 : chx Attachment: http://drupal.org/files/issues/fix_form_0.patch (4.72 KB) Small bugfix. ------------------------------------------------------------------------ Sat, 23 Apr 2005 14:53:05 +0000 : chx doh. still has serious problems when the name is an array. stay tuned. ------------------------------------------------------------------------ Sun, 24 Apr 2005 09:13:00 +0000 : chx Attachment: http://drupal.org/files/issues/fix_form_1.patch (4.97 KB) Recursive version. ------------------------------------------------------------------------ Sun, 24 Apr 2005 12:51:18 +0000 : Dries Patch looks good but I would add some phpdoc to document what and why we are doing this. ------------------------------------------------------------------------ Sat, 21 May 2005 16:43:47 +0000 : chx Attachment: http://drupal.org/files/issues/fix_form_2.patch (3.23 KB) 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. ------------------------------------------------------------------------ Sat, 21 May 2005 16:55:24 +0000 : chx Attachment: http://drupal.org/files/issues/fix_form_3.patch (3.24 KB) ------------------------------------------------------------------------ Sat, 21 May 2005 18:34:20 +0000 : Dries Committed to HEAD. Thanks. ------------------------------------------------------------------------ Mon, 23 May 2005 00:09:30 +0000 : TDobes 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. ------------------------------------------------------------------------ Tue, 06 Sep 2005 03:00:14 +0000 : mr700 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 [1]) still applies with little fuzziness and fixes the problem. [1] http://drupal.org/node/18663#comment-30411

1 0

[drupal-devel] callback arguments
by Gerhard Killesreiter 08 Sep '05

08 Sep '05

Hi there, today Dries has a committed a patch that removes the last occurrence of a non-array menu callback argument from core. It is actually documented that those should be arrays, but in php4 it also works with strings. The following contrib modules need fixing: banner/banner.module: cck/content.module: contact_dir/contact_dir.module: diff/diff.module: htmlarea/htmlarea.module: libdb/libdb.module: paypal_tipjar/paypal_tipjar.module: pdfview/pdfview.module: periodical/periodical.module: quiz/quiz.module: role_to_file/role_to_file.module: stats_plot/stats_plot.module: trackback/trackback.module: video/video.module: weblink/weblink.module: Cheers, Gerhard

1 0

[drupal-devel] [bug] Block table: no key/index defined
by Souvent22 07 Sep '05

07 Sep '05

Issue status update for http://drupal.org/node/27157 Post a follow up: http://drupal.org/project/comments/add/27157 Project: Drupal Version: cvs Component: database system Category: bug reports Priority: normal Assigned to: Anonymous Reported by: m3avrck Updated by: Souvent22 Status: patch (code needs review) I agree taht some sort of indexing needs to be done on the blocks. Although, on "most" sites, this is not a problem. But, with any scale (be it many many blocks, or many many hits), the blocks will be retrived either often (site with many hits) or become a large query (many blocks, however if you have that many blocks, you may want to look at your site design). Either way, this would make for a performance improvement. I know when optimizing a site, especially one that recieves many hits; it's hard to find just one improvement, it's usually the little tweaks hear and there that result in an overall large perfomance gain. Souvent22 Previous comments: ------------------------------------------------------------------------ Wed, 20 Jul 2005 18:34:30 +0000 : m3avrck The block table currently has no primary key or index defined. This greatly hurts queries on the table, as the SQL server has no way to introduce fast, index, based searching, which can be cachable. More info: http://www.odetocode.com/Articles/237.aspx I recommend the following change: PRIMARY KEY(module,delta); That should satisfy the key requirements for the table and speed up queries on this table, with most noticeable results on websites that rely on lots of modules. Fix should be backwards compatible as well. I'd post a patch for this database change but I'm not sure the exact procedures for posting a database change patch, not as straightforward, so if anyone cares, that would be great. Nice little, almost free performance benefit for the next HEAD version :-) ------------------------------------------------------------------------ Sat, 23 Jul 2005 10:28:56 +0000 : djnz Disagree. Blocks are infrequently retrieved from the table by module and delta. In fact the only query that is run on every page is SELECT * FROM {blocks} WHERE status = 1 AND region IN ('%s') ORDER BY weight, module ... so status and region would be the first things to index (unless you have a LOT of blocks active, weight and module are not going to make any significant impact). ------------------------------------------------------------------------ Mon, 25 Jul 2005 02:08:29 +0000 : m3avrck Ok I do agree with you and maybe my post wasn't super optimized. There should be some sort of *unique* key on the table and drumm agreed with me in #drupal that PRIMARY KEY(module,delta) would be a great candidate. As for the indexes, we can index the status and region since those are fields that queries actually act upon. But we should have definetly some sort of unique key defined as well for the best optimization, along with an index on most queried fields. ------------------------------------------------------------------------ Wed, 07 Sep 2005 14:48:56 +0000 : m3avrck Going to bump this, I think we can get this performance fix into 4.7. I would suggest based on comments: PRIMARY KEY(module,delta);<br> INDEX(status,region); If anything, the PRIMARY KEY could be left off, however I see this as a small additional performance boost... only hinderance on performace would be on insert new blocks... which is significantly less frequent than actually showing/displaying blocks. ------------------------------------------------------------------------ Wed, 07 Sep 2005 15:23:15 +0000 : m3avrck Attachment: http://drupal.org/files/issues/database_2.patch (2.46 KB) Here's a patch that adds this support to the latest HEAD. Tested on MySQL, both the update patch works and dropping/recreating database patch works. Comments welcome on logistics of key/index, but otherwise should be a nice little performance boost for 4.7. ------------------------------------------------------------------------ Wed, 07 Sep 2005 21:51:13 +0000 : m3avrck Will reroll patch tomorrow (seeing lots of updates going in right now).

1 0

[drupal-devel] [bug] Block table: no key/index defined
by m3avrck 07 Sep '05

07 Sep '05

Issue status update for http://drupal.org/node/27157 Post a follow up: http://drupal.org/project/comments/add/27157 Project: Drupal Version: cvs Component: database system Category: bug reports Priority: normal Assigned to: Anonymous Reported by: m3avrck Updated by: m3avrck Status: patch (code needs review) Will reroll patch tomorrow (seeing lots of updates going in right now). m3avrck Previous comments: ------------------------------------------------------------------------ Wed, 20 Jul 2005 18:34:30 +0000 : m3avrck The block table currently has no primary key or index defined. This greatly hurts queries on the table, as the SQL server has no way to introduce fast, index, based searching, which can be cachable. More info: http://www.odetocode.com/Articles/237.aspx I recommend the following change: PRIMARY KEY(module,delta); That should satisfy the key requirements for the table and speed up queries on this table, with most noticeable results on websites that rely on lots of modules. Fix should be backwards compatible as well. I'd post a patch for this database change but I'm not sure the exact procedures for posting a database change patch, not as straightforward, so if anyone cares, that would be great. Nice little, almost free performance benefit for the next HEAD version :-) ------------------------------------------------------------------------ Sat, 23 Jul 2005 10:28:56 +0000 : djnz Disagree. Blocks are infrequently retrieved from the table by module and delta. In fact the only query that is run on every page is SELECT * FROM {blocks} WHERE status = 1 AND region IN ('%s') ORDER BY weight, module ... so status and region would be the first things to index (unless you have a LOT of blocks active, weight and module are not going to make any significant impact). ------------------------------------------------------------------------ Mon, 25 Jul 2005 02:08:29 +0000 : m3avrck Ok I do agree with you and maybe my post wasn't super optimized. There should be some sort of *unique* key on the table and drumm agreed with me in #drupal that PRIMARY KEY(module,delta) would be a great candidate. As for the indexes, we can index the status and region since those are fields that queries actually act upon. But we should have definetly some sort of unique key defined as well for the best optimization, along with an index on most queried fields. ------------------------------------------------------------------------ Wed, 07 Sep 2005 14:48:56 +0000 : m3avrck Going to bump this, I think we can get this performance fix into 4.7. I would suggest based on comments: PRIMARY KEY(module,delta);<br> INDEX(status,region); If anything, the PRIMARY KEY could be left off, however I see this as a small additional performance boost... only hinderance on performace would be on insert new blocks... which is significantly less frequent than actually showing/displaying blocks. ------------------------------------------------------------------------ Wed, 07 Sep 2005 15:23:15 +0000 : m3avrck Attachment: http://drupal.org/files/issues/database_2.patch (2.46 KB) Here's a patch that adds this support to the latest HEAD. Tested on MySQL, both the update patch works and dropping/recreating database patch works. Comments welcome on logistics of key/index, but otherwise should be a nice little performance boost for 4.7.

1 0

[drupal-devel] [bug] Can't delete subjects with '&' in title
by Cvbge 07 Sep '05

07 Sep '05

Issue status update for http://drupal.org/node/27140 Post a follow up: http://drupal.org/project/comments/add/27140 Project: Drupal Version: cvs Component: contact.module Category: bug reports Priority: normal Assigned to: Anonymous Reported by: m3avrck Updated by: Cvbge Status: patch (code needs review) Hi, patch does not applies to currect cvs anymore. For the database.pgsql you need to remove (11) from int(11) and also first "category" from UNIQUE line, so it looks like this: CREATE TABLE contact ( cid int NOT NULL, category varchar(255) NOT NULL default '', recipients text NOT NULL default '', reply text NOT NULL default '', UNIQUE (category), PRIMARY KEY (cid) ); The updates are also not compatibile with postgres, but postgres upgrade path is completly broken in cvs and I'll fix it later, so do not pay attention to it. Cvbge Previous comments: ------------------------------------------------------------------------ Wed, 20 Jul 2005 14:57:03 +0000 : m3avrck Using the site-wide contact forms, any subjects with an '&' cannot be deleted. Clicking delete for these subjects brings up the expected action flow, however, confirming the delete does nothing. Subject still exists. ------------------------------------------------------------------------ Mon, 25 Jul 2005 13:45:38 +0000 : m3avrck It appears this issue might be linked with this one: http://drupal.org/node/23685 A problem with mod_rewrite and not the actual contact module. ------------------------------------------------------------------------ Wed, 31 Aug 2005 16:51:56 +0000 : Souvent22 Attachment: http://drupal.org/files/issues/contact.module_6.patch (4.7 KB) Redid how contacts work. It now references contacts by ID instead of their category as the unique identifier. It also allows for contacts to have weights, so that one may control where the contact will appear in the listing. ------------------------------------------------------------------------ Wed, 31 Aug 2005 16:53:01 +0000 : Souvent22 Attachment: http://drupal.org/files/issues/updates.inc (27.23 KB) This is the update.inc for the database cahnges that goes along with this patch. ------------------------------------------------------------------------ Wed, 31 Aug 2005 18:31:37 +0000 : Souvent22 Attachment: http://drupal.org/files/issues/contact.module_7.patch (4.5 KB) Updated patch. ------------------------------------------------------------------------ Wed, 31 Aug 2005 18:31:55 +0000 : Souvent22 Attachment: http://drupal.org/files/issues/updates.inc_4.patch (1.29 KB) Updated database patch. ------------------------------------------------------------------------ Sat, 03 Sep 2005 16:17:21 +0000 : Cvbge Hi, 0. You should keep all changes in one file, i.e. pake one patch, not patch for every file 1. You should include changes for database/database.{mysql,pgsql} 2. IIRC auto_increment was depreciated, you should use db_next_id() (but I see auto_increment in cvs...) 3. You use "WHERE cid = '%s'", but cid is integer so this should probably be %d (although this probably would not break anything..) ------------------------------------------------------------------------ Wed, 07 Sep 2005 17:31:21 +0000 : Souvent22 Attachment: http://drupal.org/files/issues/contact_full.patch (6.38 KB) Ok, I've taken your suggestions in, and i think I have it this time. Let me know if this is a correct format for a patch. This only fixes the bug of not being able to delete special characters. ------------------------------------------------------------------------ Wed, 07 Sep 2005 17:37:30 +0000 : m3avrck Check your database updates, the UNIQUE fields don't look right syntactically. ------------------------------------------------------------------------ Wed, 07 Sep 2005 18:23:07 +0000 : Souvent22 Attachment: http://drupal.org/files/issues/contact_full_0.patch (6.25 KB) Ok, i apologize for the last upload. Pleasle ignore. I send the wrong file. :-[. Anyway, here's a tested, tried, and correct patch. ------------------------------------------------------------------------ Wed, 07 Sep 2005 18:26:33 +0000 : m3avrck +1 patch works great! Latest HEAD version applied with update.php works great, modifies database correctly for MySQL (and PostgreSQL looks correct as well) and contacts can now be deleted if they have '&' in the subject. Also, dumped entire database and created from modified database.mysql and everything works great. I'd say this patch is ready to be committed!

1 0

[drupal-devel] [bug] NOTICES silenced
by hlslaughter 07 Sep '05

07 Sep '05

Issue status update for http://drupal.org/node/30633 Post a follow up: http://drupal.org/project/comments/add/30633 Project: Drupal Version: cvs Component: base system Category: bug reports Priority: normal Assigned to: Anonymous Reported by: hlslaughter Updated by: hlslaughter Status: patch (code needs review) Attachment: http://drupal.org/files/issues/common.inc_hlslaughter_200509071421.patch (1.88 KB) this does not necessarily silence all NOTICEs, just the most obvious ones that are generated by using a pretty generic HEAD drupal install. i'm not sure if i've followed coding standards correctly, so i'd appreciate feedback on even the slightest thing :) again, attaching a patch crashed mozilla, so will try to attach with followup post hlslaughter

1 1

[drupal-devel] [task] Performance: fix all include paths
by m3avrck 07 Sep '05

07 Sep '05

Issue status update for http://drupal.org/node/30122 Post a follow up: http://drupal.org/project/comments/add/30122 Project: Drupal Version: cvs Component: base system Category: tasks Priority: normal Assigned to: Anonymous Reported by: m3avrck Updated by: m3avrck Status: patch (code needs review) Attachment: http://drupal.org/files/issues/drupal_2.patch (9.68 KB) This patch fixes all of the outstanding Drupal include paths by placing the './' in front of the include path. This brings all of the Drupal core files up to date with the boostrap.inc which made first use of this performance gain. In simple testing, I've have noticed much faster load times with Drupal on my slow machine. Definetly makes a big difference! m3avrck

2 4

[drupal-devel] drupal site on forwarded server
by openwereld.net 07 Sep '05

07 Sep '05

Hi list, I entered this first as an issue on the drupal site at http://drupal.org/node/30173, but got no reaction there. It's probably more suitable for a developer's discussion. If a drupal site is situated on an intranet and external requests are forwarded to it, the function 'conf_init' in bootstrap.inc uses the 'wrong' host name. It always uses $_SERVER['HTTP_HOST'], while in case of request forwarding it could use $_SERVER['HTTP_X_FORWARDED_SERVER']. The following code fragment (in conf_init() ) should improve this: if( isset( $_SERVER['HTTP_X_FORWARDED_SERVER'] ) ) $server = explode('.', rtrim($_SERVER['HTTP_X_FORWARDED_SERVER'], '.')); else $server = explode('.', rtrim($_SERVER['HTTP_HOST'], '.')); john.

1 0

[drupal-devel] [bug] Correct progress bar height
by drumm 07 Sep '05

07 Sep '05

Issue status update for http://drupal.org/node/30629 Post a follow up: http://drupal.org/project/comments/add/30629 Project: Drupal Version: cvs Component: base system Category: bug reports Priority: minor Assigned to: drumm Reported by: drumm Updated by: drumm Status: patch (ready to be committed) Attachment: http://drupal.org/files/issues/drupal.css_2.diff (525 bytes) The height of the progress bar changed from 2em to 1.5 em at some point and the filled section is still taking up 2em. Steven agreed this fix is correct in IRC yesterday. drumm

1 0