October 2005 - Development - lists.drupal.org

[drupal-devel] Re: Request for input on Quiz module data
by David Norman 13 Oct '05

13 Oct '05

Disclaimer: I'm not totally familiar with the quiz module, but I did read the whole RFC. Regarding the new schema quiz table: 1. "number_of_questions" can be queried and therefore should not be stored IMO if you're going for full normalization. 2. I think "takes" could be have a more descriptive name and should be bigger than tinyint. drupal-devel-request(a)drupal.org wrote: > ------------------------------ > Message: 8 > Date: Wed, 12 Oct 2005 19:36:25 -0400 > From: Angie Byron <drupal-devel(a)webchick.net> > Subject: [drupal-devel] Request for input on Quiz module data > structure [long] > To: drupal-devel(a)drupal.org > Message-ID: <434D9DF9.4090309(a)webchick.net> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > Hi, folks. Hope that this is the right place for this. Apologies in advance for > the length, but I wanted to make sure everyone's on the same page before asking > for feedback. > == BACKGROUND == > As most know, I worked over the summer on the Quiz module. Thanks to a generous > injection of funding from Ejovi Nuwere at SecurityLab.net, I've been able to > both complete most of the work that was not done by the other student assigned > to this project during SoC, and am also able to now focus on adding additional > features to the module. > The quiz module was developed around Robert Douglass's specification (which I've > temporarily mirrored here: http://drupaldev.snarkles.net/node/2) In a nutshell, > this means rather than creating a simple module that could do multiple choice > questions, instead creating a quiz *framework* which could potentially support > almost limitless question types, from traditional multiple choice/fill in the > blank, to really innovative ideas such as "reading" a question aloud to gage > reading comprehension and so on. > == HOW IT CURRENTLY WORKS == > The current schema can be viewed here: > http://cvs.drupal.org/viewcvs/drupal/contributions/modules/quiz/quiz.mysql?… > For the more visually inclined, I've uploaded an E-R Diagram here: > http://www.webchick.net/soc/quiz.module/quiz-schema-old.png > This is based off the original proposal, where serialized arrays (/me ducks in > advance) are used to store answer, feedback and user response data. > NOTE: Skip to "THE PROBLEM" if you don't care about how this stuff is stored > currently :P > A typical result array (stored in quiz_results.results) might look like: > Array > ( > [10] => Array > ( > [input] => Array > ( > [0] => 0 > [1] => 2 > [2] => 3 > ) > [feedback] => Array > ( > [0] => Correct! > [1] => Correct! > [2] => Correct! > ) > [score] => 1 > ) > [6] => Array > ( > [input] => Array > ( > [0] => 2 > ) > [feedback] => Array > ( > [0] => Nope, fooled you! > ) > [score] => 0 > ) > ... > ) > The array keys (10 and 6) are question node IDs... in other words, this is how > the user responded to both question 10 and question 6. Over the course of the > quiz, this array keeps growing, with one array of results per question. > "input" is an array of the answer(s) that the user selected on that question. > The first question was a multiple choice question, so the user plugged in three > answers (answers 1, 3, and 4 -- remember that arrays start at 0 index). The > second question was only a single-choice question, and the user selected answer 3. > "feedback" is an array of the feedback to the answer(s) the user selected for > input. As you can see, the user answered correctly on all three answers on the > first question, but did not answer correctly on the second question. > Finally, "score" is either a 1 (indicating the answer was correct) or a 0 > (indicating the answer was not correct). > == THE PROBLEM == > The question was raised to me: > "If there's a million questions, and a > million students, and we want to find out for a given > question, how many student got it right how do we do that?" > Right now, the answer is something along the lines of this: > ----------------------------------------------------------------------- > function get_question_results($quiz_nid, $question_nid) { > // Array to store question results > $question_results = array( > 'right' => 0, > 'wrong' => 0 > ); > // Grab a list of the results array for each quiz attempt > $quiz_results = array(); > $results = db_query("SELECT r.results FROM {quiz_results} r WHERE r.quiz_nid = > %d", $quiz_nid); > while ($result = db_fetch_object($results)) { > $quiz_results[] = unserialize($result->results); > } > // Now, loop through each result and look for the given question > foreach ($quiz_results as $key => $value) { > if ($value[$question_nid]) { > // If question is found check if it was answered correctly or not > if ($value[$question_nid]['score'] == 1) { > $question_results['right']++; > } else { > $question_results['wrong']++; > } > } > } > return $question_results; > } > // Call the function and output results > $results = get_question_results(12, 10); > echo $results['right'] ." students answered this question right, " . > $results['wrong'] ." answered it wrong."; > ----------------------------------------------------------------------- > This is quite processor and RAM-intensive, since it has to load ALL of the > results for each attempt on a given quiz first, then parse through those to find > individual question attempts, and then finally retrieve the answers from each > and evaluate whether or not they were correct. > And this same basic logic has to be completed for each potential query someone > might want to throw at it.. Such as, "What is the average score of this user > among all exams?" > In short, it doesn't look like this solution is going to be very scalable in the > long-term for the kind of ad-hoc querying that educators are likely going to > really want to be able to do. > == THE SOLUTION? == > I've created an alternate, normalized data schema and wanted to get your > feedback on it, if you get a chance. > http://webchick.net/soc/quiz.module/newschema.sql > E-R Diagram: http://webchick.net/soc/quiz.module/quiz-schema-new.sql > Changes: > - quiz/quiz_questions tables stay the same > - multichoice turns into 'quiz_question' containing only a primary key and a > "properties" field to take the place of "multiple_answers." I envision this > field as being more flexible than multiple_answers, as it could potentially > store a plethora of options for very diverse array question types, perhaps in > the form of key/value pairs or something (multichoice=1;email_results=1; etc.) > - answers are now broken out into a 'quiz_question_answer' table, which links > back to quiz_question with separate fields for answer, feedback, and points. The > answer field itself could store a wide variety of options, such as "Answer" for > a multichoice question, "St. Paul|Saint Paul" for a fill-in-the-blank question, > and so on. Since answer display is handled in each question type module in the > _form and _render_question hooks, we can still have the advantage of storing > data independently from the actual quiz module. Feedback, too, can be stored in > whatever way makes sense for the answer, and points can indicate whether a > question is incorrect (0 points) or can also be used for 'weighting' questions > (answer 1 is 20 points, answer 2 is 50 points, etc.) > - Results are now stored in two different tables: quiz_result for the overall > properties (start time/end time, etc. -- essentially the same as the current > table) and quiz_question_results which provides the textual representation of > what that user put for his/her answer on each question attempt. This could be > stored in a similar manner to the answers table... for multiple answer > questions, it could be 23,63,3 to indicate the answer ids selected, for > fill-in-the-blank it could have 'Minneapolis', etc. Each question type's > _evaluate_question hook would take the data from the answer field and check it > against the given question's answers. > - Fields representing times and dates are switched to DATETIME instead of > INTEGER (for UNIX timestamp) in order to help with date range calculation > And if you made it through that, a hearty congratulations. :P > Any thoughts? > ------------------------------

1 0

[drupal-devel] SQLite Support
by Sergio 12 Oct '05

12 Oct '05

Can someone help me here? I was trying to use these on drupal 4.6.3 but I don't know why it always give me 'access denied' to the front page... Any enlightenment? Thanks, - Sergio

2 1

Re: [drupal-devel] permissions page
by Kobus Myburgh 12 Oct '05

12 Oct '05

Not really sure how this will improve this problem. Besides, iframe is not cross-browser, is it? Isn't iframe a Netscape thing? Kobus >>> dopry(a)thing.net 10/11/2005 8:49 PM >>> Is there a reason to not just render the permissions table in a horizontally scrollable iframe? Permission Group +------------------------------------+ | role1 role2 role3 | Permission 1 | | Permission 2 | | Permission 3 | | |<-----------------------##--------->| +------------------------------------+ On Fri, 2005-10-07 at 17:09 -0400, Herman Webley wrote: > Your solution doesn't solve the specific problem that Karoly is trying > to address. That issue I believe is the space consumed by the current > iteration of the permissions page. > > What do you see as the scalability issues related to Karoly's proposed solution? > > Your solution does sound good though. It would require a lot of > scrolling, but thats the only problem. > > On 10/7/05, Gerhard Killesreiter <killesreiter(a)physik.uni-freiburg.de> wrote: > > > I don't like the idea of having text as images. Doesn't scale. Also, if > > you scroll down onthat page, you still won't know to which role you are > > assigning perms. I think my solution is simple, accessible, and solves all > > problems. > > > > Cheers, > > Gerhard > > > > > -- > Best regards, > Herman Webley > >

3 2

Re: [drupal-devel] permissions page
by Kobus Myburgh 12 Oct '05

12 Oct '05

Ah, I stand corrected. Thanks! I briefly checked an HTML book that I had on my desk, and it attributed that to Netscape Kobus >>> web(a)timaltman.com 10/12/2005 9:11 AM >>> On Wed, 12 Oct 2005 08:24:20 +0200, Kobus Myburgh <ITBJDM(a)puknet.puk.ac.za> wrote: > Not really sure how this will improve this problem. Besides, iframe is > not cross-browser, is it? Isn't iframe a Netscape thing? No, IFRAMEs are part of the HTML 4.01 Transitional DTD. -- Tim Altman

1 0

Re: [drupal-devel] permissions page
by Kobus Myburgh 11 Oct '05

11 Oct '05

I am not sure whether I understand your proposal correctly. See attached mock-up to see if I understand you correctly. If I do, my first idea here won't be feasible... First idea: This probably doesn't belong in this post, but I am posting here because it seems a refactoring of this page is eminent. While this is the case, it should also be done that the header row is repeated every 20 lines or so, because if you have a lot of modules installed, and a few roles (any amount of roles that require you to scroll sideways to edit) makes it very tedious to assign permissions. Second Idea: Since performance is not really an issue here, why not have different pages for each role (selectable by a drop-down, perhaps)? That way you can set the permissions for one user at a time, but you can set all modules in one sitting. If you have few roles and few modules, it will be slightly more work than now, but if you have a lot of modules/roles, this would be a great change. Third Idea: Add a checkbox "All" which will give ALL roles the specific permission. Regards, Kobus >>> karoly(a)negyesi.net 10/7/2005 3:35:41 PM >>> Role names take too much place and vertical text is only in CSS3 (and in IE for long) So, I thought of diagonal text, think one table cell per letter. There are javascripts for this out there, it would be very easy to do in PHP. Only one (or few?) modules would be displayed at once. To make the presumable huge table fit, let's make it unthemed with just a link to 'admin' to save space. Color the table like the chess board. Maybe 3 x 3 checkboxes would be one color. Regards NK

21 42

Re: [drupal-devel] Improvements abandoned? (was: bug squashing session)
by Steve Dondley 11 Oct '05

11 Oct '05

What happened to having the bug reports and updates getting mailed to this mailing list? That seemed to work. I posted three bug fix patches and filed two bug reports yesterday but I feel like they are left in a black hole. That's not a feeling you want to leave folks with after they contributed several hours of their time. I know you can easily see the list at http://drupal.org/project/issues?projects=3060&states=1&categories=bug. But you know the saying, "out of sight, out of mind." Maybe getting bug reports sent to this mailing list again would help make things along. On 10/11/05, Richard Archer <drupal.org(a)juggernaut.com.au> wrote: > At 12:04 PM +0200 11/10/05, Gerhard Killesreiter wrote: > > >Since there will be a lot of developers the bug list can get lengthy, > >don't be shy. > > This is an excellent plan and will go a long way towards > shortening the backlog of bugs as well as improving the quality > of Drupal. > > But there still remains the problems of: > > - all the feature requests and tasks in the queue > > - the fact that the queue grows faster than it is processed > > - the way unwanted patches just sit in the queue forever > without being reviewed and rejected. > > A longer term process needs to be established to make sure worthy > patches are considered and also that reviewers/contributors don't > waste their time. > > ...Richard. > -- Dondley Communications http://www.dondleycommunications.com Communicate or Die: American Labor Unions and the Internet http://www.communicateordie.com

4 4

[drupal-devel] couple of SQL questions
by Gordon Heydon 11 Oct '05

11 Oct '05

Hi, I have a couple of issues that I have been playing with for a project and can't quite get my head around it from a SQL POV. Is it possible to list all the rows in a table where one column has more than one occurrence of the same value. In this can I am storing the ip address of some submitted data, and I need to check to see if they have submitted more than 1 row. I don't care about the ones with a single occurance, but I need to know about the ones that appear more that one. My second question is that. I have 2 tables, one is a teams table which has a list of entered teams. The other table is a list of members for each team. One of the columns is the name of the member. What I would like to do is create a select from the team table which has a single column with a comma separated list of members in a single field. This means I could sort and use it as a method of detecting entries of multiple teams with the same members. Or if anyone has a better ideas on how to detect duplicate entries would be most appreciated. Thanks in advance. Gordon.

4 6

[drupal-devel] Improvements abandoned?
by Steve Dondley 11 Oct '05

11 Oct '05

I submitted a couple of patches a few months ago to improve some of Drupal's help text. See: http://drupal.org/node/31394 http://drupal.org/node/27073 Why the delay? These patches probably won't even apply now. Please hit me with a clue bat if my complaint here is out of line. -- Dondley Communications http://www.dondleycommunications.com Communicate or Die: American Labor Unions and the Internet http://www.communicateordie.com

4 5

[drupal-devel] usability: a results oriented user interface, good read
by Theodore Serbinski 11 Oct '05

11 Oct '05

Very interesting: http://www.useit.com/alertbox/wysiwyg.html Maybe we can use this concept in future Drupal versions. "I want to administer a site" "I want to create content" "I want to add pictures to my content" "I want to..." And so forth. ted (aka m3avrck)

3 2

[drupal-devel] multidimensional arrays VS objects. should we go for one?
by Bèr Kessels 10 Oct '05

10 Oct '05

Hi, When designing the patch for hook_links to return structured data, instead of that accursed string, I found that we have no "good practice" nor consistency in core when to use objects and when to use keyed arrays. So, Is it an idea to discuss this at druopalCON, to discuss this here? Or does it not really matter to others? I would prefer to have One Method. using structured arrays would mean converting all(!) $node->foo (and the likes) stuff to $node['foo']. While using the other one would probably mean changing all (haha) form api stuff to objects, but certainly all menu items. I think it is doable like it is now, but that it is just Not Very Clean. (and will result in problems, like we had for php5) IMO the clean way is to have arrays when the data is NOT structured, but is a list of similar items, but to have objects when it is structured. so $menu_item->path instead of $menu_item['path']. just like we have for most 'things' like nodes, comments, terms, etc. But all these $menu_items should sit in a $items array, not in $items->item1, $items->item2. Most of this is in place already, but not all. Keeping Drupal clean means deciding on standards like these too, IMO. And oh, I am not saying that we should rewrite all this immediately, but enforcing such guidelines/standards form now on will make sure we don not introduce more inconsistancy. So, opinions?

5 4