[DISCONTINUED] Wish-list for MATLAB Answer sections.
Walter Roberson
on 8 Feb 2011
Latest activity Reply by Walter Roberson
on 12 Feb 2023
This topic (which was not started by Mathworks) is for features you would like to see for this MATLAB Answers facility.
I suggest one wish per answer, so that people can vote for individual wishes.
What should you post where?
Next Gen threads (#1): features that would break compatibility with previous versions, but would be nice to have
@anyone posting a new thread when the last one gets too large (about 50 answers seems a reasonable limit per thread), please update this list in all last threads. (if you don't have editing privileges, just post a comment asking someone to do the edit)
578 Comments
Time Descending
This particular Question is for discussion changes to the MATLAB Answers software itself -- the forum software that allows people to ask questions and make responses.
You should have created a new Question for that topic.
estimate the current height change along the course between point A=(sqrt3/2,1/2) and B=(1/2,sqrt3/2) identitfy 2 ways to estimate this height change one with derivatives one without
power = 2;
f = @(x, y) x .* y.^power;
A = [sqrt(3)/2, 1/2];
B = [1/2, sqrt(3)/2];
t = linspace(0, 1, 1000);
x = (1 - t) * A(1) + t * B(1);
y = (1 - t) * A(2) + t * B(2);
z = f(x, y);
height_change = z(end) - z(1);
disp(height_change);
this gave the answer of 0.1585 but it doesnt seem correct if i change the ^ to * the answer becomes 0 am i correctly computing what the question is asking
Dark theme/mode?
that already exists for MATLAB Answers. Go to the top right of any code block in MATLAB Answers, and click on the half filled circle, and you will toggle dark mode.
%example code block
Reminder: this Question is about the MATLAB Answers facility (this forum), with the MATLAB product wishlists being in other Questions.
Users need acces to eig function
Lapack is called in most circumstances where it makes sense to do so. Generally speaking, MATLAB delegates most vector and array mathematical operations to high-speed multithreaded libraries provided that the amount of work to be done is large enough to make it worth going through the interface code.
Yes, but why do you ask? Is there a particular reason you need or want to know what the internal implementation is doing? Why do you care as long as you receive the right answer?
Thank you Walter and Steven for providing valuable comments on this topic, precisely for what Walter said I decided to mention this as a wish in the wish-list and not as a question.
Do you know of any other kind of funstions in Matlab that migh be calling LAPACK or other libraries outside Matlab?
Thank you again
This is not going to happen, for performance reasons. If I scan the eigs code correctly, the circumstances under which eig is invoked would be cases where eig would need to use the qz algorithm, which it does by calling lapack. There is no hidden matlab-language source code for eig, to be stepped through.
I am specially interested in accesing the imaginary part of the eigensolution. However, when using eig, I can not "step in (F11)" into eig.m
For this reason I implemented a QR-based eigensolver in which it is possible for me to acces the information. But I think, it would be nice that eig.m could be executed line by line, just like eigs.m for instance. That way, the user can have more control on the flow of the information
https://www.mathworks.com/matlabcentral/fileexchange/16777-lapack provides an interface to call lapack directly
http://www.netlib.org/utk/people/JackDongarra/etemplates/node282.html
What "partial results" are you interested in observing? Note that if you have a particular algorithm in mind for computing eigenvalues and eigenvectors that takes a certain series of steps, the eig function's code may not use that same algorithm and may not need to compute those same intermediary results.
Since this isn't related to MATLAB Answers but to MATLAB in general you might want to ask it as a separate question and/or send an enhancement request for eig to return the additional information you seek to Technical Support. If you submit an enhancement request to Support a description of how you plan to use this information if it becomes available would be useful to provide some context for the request.
If you want to test that the eig function returns correct answers, on the other hand, you could always write your own tests. We have thorough testing for the eig function but I can certainly understand the "trust but verify" mindset.
If you open eigs.m, you will see it calls eig.m in lines 1026:1038, now, if for any reason you need to verify what eig., does, you can not, probably Matlab is calling LAPACK to solve eigenvalue problems with eig.m
But as a user, it would be nice if I can verify partial results during the calculation of eigenvalues and eigenvectors.
In my particular case, I needed eigenvalues, but also additional information handle by the eigensolver that was not available with matlab.
Could you expand on this?
Option to bookmark an answer or the whole question thread. And this should be available to us in the profile dashboard. I bookmark the page right now. Thanks for this thread too!
I often also store question and the accpeted answer in my MATLAB code library. I save them with a CodLib_typeOfAction filename format so I know what is in there. So CodLib_Extraction would be a live script with several sections , each with extraction example and with my comments. I was initially weak with extraction in MATLAB but now I am good with it. I run it and test these Q&As before saving so in future its good to go. Plus it helps me breakdown the problem statement and understand better.
Browser Bookmarks are normally when I find new functions. or an unique problem type.
Have not used Trello yet. Will check it out! thanks. I also think spell check for these answer boxes will be great!
There's lots of way to use trello. I use it as a project management tool and as a means of collect ideas.
What is Trello? (Link?) I don't use Chrome as my preferred browser because Chrome and Edge won't let you pin your heirarchical bookmarks panel to the left side of the screen like Firefox will. That's a deal breaker for me.
I used to use Chrome's bookmark tool so I could nest URLs by topic in subfolders. For example. Matlab > Plots, Matlab > fitting, etc.
But after accumulating may bookmarks over the years, I switched to using a Trello board which gives even more options for organization.
I'm sure you're aware of these options already. I think they are better than an internal bookmark system. I wonder if access to those internally stored bookmarks would still be available if a user ever stopped their subscription.
The ability for editors/moderators to split and combine threads.
- A comment is actually an answer, so being able to change it to being an answer
- The OP replies to a question by creating an answer, so being able to change an answer to a comment of the selected thread.
This was the highest rated request during a recent internal process.
For easy navigation and for the historians,
Anyone with edit privileges should feel free to update this whenever needed.
Well, after the last update of Matlab Answers my biggest wish is now to put the white space back between the answers! I don't know if this was a common request to get rid of it, but it looks dreadful with answers just all running into each other. I like white space in my code and I like it in an answers forum for the same reasons. It's unreadable without. Distinct answers should be properly delineated, rather than just a thin box around them all slammed together.
I don't like that copy-pasting a user's handle only copies the text rather than copying it as a link. Additionally, when you paste the handle, it adds a line break prior to the text (try copy-pasting any user handle here).
What I'd really like to see is a spellcheck option. I realize variable names would often be detected as misspellings but that would serve as an additional reminder to format code (which would then be ignored by the spell checker).
As far as I can see, Activity Feed / Manage Followed Content apparently only shows threads that I follow if there is a post to that thread after I follow it. That is, if I follow a thread but no one later posts to it, the thread does not show up in my Followed Content list.
I'd like to have all of my followed threads show up in that list, even if there has been no further activity. This is because I might want to post a suggestion or partial answer later, if the question remains unanswered.
Alternatively, could there be a way to filter by Unanswered+Followed?
(Answers Dev) I will forward your message on to the appropriate developer.
Do we have to have the 'This website uses cookies...' message at the bottom of the Matlab Answers website every single day when we navigate to the page? I know the internet is now festooned with these kinds of messages, but it's not as though this one is giving options to select, it is just information. I've been signed into my account the whole time - I know nothing about website design, but can it not realise that if you are logged in you have already seen the message?!
Is it worth to elaborate an old answer that's been accepted? Will anybody ever read it?
 
Wish:   A symbol to the left of the question, which indicates whether the question is on a "watch-list".
Problem: Many users do not understand how to use the markup button, {}Code, which results in formatting of text like the example below
.
if true
Error using assignin Unknown command option.
Error in cssm_c (line 20) assignin( 'base' , col_names(i) , num(:,i) );
end
.
Proposal: "{}Code" shall check whether any lines are selected and if not issue a warning. Or something better.
I used to do that for images until I learned that it behaves differently if you do that (put in angle brackets yourself) than if you use the image icon. Using the image icon will download the image to MATLAB Central and be permanently available, while if you do the HTML yourself, it points to the third party web site, which means the image may not be there forever.
I still use angle brackets for links because that acts the same way and it's faster.
Personally I type in all markup manually, except for doing block {} or (rarely) using the new paperclip.
What I would use would be a button that put spaces around operators and spaces after commas and semi-colons. SO that the result would flow nicely when auto-formatted. I have wasted so much time copying Maple non-spaced expressions and carefully going through them adding spacing so that they become readable.
My typical workflow is
- copy&paste code
- select code
- push {}Code
- and a little bit of fixing
I never see
if true
% code
end
Funny thing is that I don't use the buttons, B , I , and Aa in the same way, i.e. select followed by push. From now on I will use these buttons more.
I have the impression that some posters who try to markup correctly fail because they misunderstand the use of the button, {}Code.
Yeah, I think it assumes if you type that without already pasting in your code that you're just going to start hand typing your code it, which is almost never the case. Even if it were the case, posters leave the if true in there and it's just a mess, with bad indenting and often chunks of the code not formatted properly.
I agree. Perhaps something similar to add a link/picture/attachment. A popup box drops down where one can paste or write the Code and insert into the main text. if text in the main box is highlighted convert highlighted text into code.
*A modest wish:*
- The tag "oop" should qualify the post for the Category: Object Oriented Programming.
New list has been created! http://www.mathworks.com/matlabcentral/answers/42413-wish-list-2-and-bug-reports-for-matlab-answers-itself
The time-stamps for the activities log and the flagged content log appear to be in GMT. This is inconsistent with the other time-stamps that the user sees and so can lead to some confusion.
Not a major problem ;-)
I have been thinking myself that we should start a volume 2. I would not delete anything here, though.
Ideally some of what is here should be moved to the new list. I did not want to just go ahead and create the new Answers, though, as that would not give proper credit to the people who originally proposed them.
I like Daniel's Idea about leaving this alone and starting a wish-list 2.0. I can't even load this whole page without IE8 dumping. Thus any new wishes from today I can't see. Here's my first:
-Once you hit a reputation of say 20, get rid of the "how to post an answer" comments that take up space.
Now that so many of the wishes have been implemented (thank you to TMW), should we deleted the implemented answers to make it easier to find the outstanding issues? Should we refer to the current version of answers as 2.0 and start a new wish-list for it?
The new layout is a brute waste of space. The information density is sometimes lower than 3 sentences per screen.
The new boxes around everything confuse my eyes.
As of today, the new layout is wider! :)
Answers list in My Answers is sorted by old-to-new. It takes long time to go through the all answers and reach the latest one.
Please let me know if any sorting option exist to sort the Question list or Answer list. I have not found any such option ?
This regression should be fixed now.
We're aware of this regression (which somehow we manage to introduce seasonally...), and will get a fix out ASAP
Yes, TAB, I can confirm this. And it takes ages to find a list of my last ten answers. It is necessary to let the users select a sorting order and to allow a faster navigation as a popup to move 2, 4, 8, 16 pages to front or back. I hestitate to vote this question, because it cannot be found afterwards anymore in this giantic thread.
Actually I am talking about the list which shows my all the answers. Please see here
The oldest answered thread is on the top.
When I edit the Question at http://www.mathworks.com/matlabcentral/answers/42190-multiple-y-axis to reformat it, and Save, then the Save button flashes to Saving... and then reverts back to Save without the changes being saved.
Is a minimal title length of 4 words useful?
Yes, that worked. Perhaps it was the title length that was the problem.
When you have a chance could you try your edit again and see if it works? Thanks!
I was not changing the title, just the body.
Were you changing the title as well as the body of the question? We introduced a requirement that titles contain at least four "words" (whitespace-separated strings). It's possible that this validation was failing and the error message was not showing up. I'll look into it.
This Question is now so slow to load that Firefox and Safari now time-out the javascript 3 to 6 times for me.
We're aware of this issue and are investigating it.
Before today's changes, if one was editing and one highlighted text and clicked on the Code button, then after the operation was done, the selection was removed.
As of today's changes, the selection goes away for a small number of seconds while the reformat is done, but then the selection color comes back, highlighting a wider but shorter area, not covering all of the height of the original selection but covering wider (possibly even some areas that have no input.)
I noticed another weird thing about Answers. Not sure how to classify it. Maybe it's a bug, or just an unexpected/unwanted feature. <http://i46.tinypic.com/qs4dpt.jpg>
If you're creating an posting and switch to the "Preview" panel and double click a word there to highlight it, and then click some of the icons, like Bold or Italic, unpredictable things happen back in the test editor panel(not the Preview panel). Sometimes it will italicize the entire preview panel(not just the highlighted word), sometimes it inserts "_ITALICTEXT_" at random locations in the edit panel, or it might append "_BOLDTEXT_" to the text in the edit panel, sometimes it puts stars around all the text in both the edit panel and the Preview panel (but the text doesn't appear as bold though). It doesn't seem predictable. Granted, it's not something one should normally do, but I just thought I'd note it in case anyone ever decided to make some improvements to this forum, and it was easy to "fix".
Pages seem to be taking longer to load. This page especially.
With the more complicated scripts running the pages, the Answer box or the Comment box might not yet have had their style-sheet javascript loaded, so they can appear as fairly small input boxes with no controls, until (possibly minutes later) the js loading catches up and re-boxes everything.
Yes, that's what I'm using. And after the June 27 updates, more funny things started happening. Like the comment box is not as wide and then if you widen it, you can't click to place the cursor in a spot that is past a certain width. However that seems to come and go - it's not permanent. Most annoyingly, the vertical scroll bar only seems to function in the upper 75% of it's range. Clicking in the bottom 25% of the scroll bar, on the thumb bar, or on the little down arrow at the bottom of the scroll bar seem to not function at all for me. Luckily the mouse wheel still works down there so I can use that to scroll.
I confirm the funny behavior on FireFox, Win7.
To illustrate, type this in into the edit panel: word1 word2 word3. Now double click word2 in the Preview panel, not the edit panel. Now click the B or the I icon. What do you see? Why do you see any difference at all? Try it several times using B and I icons on various words. Does the text appear in the same place all the time or where you'd expect it (No). It might add BOLD TEXT after the word 3 while other times it will act on word 2 like it should, or even insert BOLD TEXT in between word2 and word 3 instead of appending it after word3 like it did at first.
In my experience, if one clicks on the Bold or Italic button when one does not have anything in the input area selected, then the _ITALICTEXT_ or *BOLDTEXT* is inserted.
Bold text is only bolded if it is within the same paragraph (and then only if there is whitespace before the first asterisk and after the last asterisk), so if you happen to have an empty line selected when you click on the bold icon, the * will be inserted but nothing will be bolded in the output.
Hey everyone, I propose that there should be a built in way to add attachments such as images. I think it would be beneficial when code is attached (such as providing the formula, expected plot to be generated etc.)
Attached code is less useful than embedded code.
This is effectively the same as http://www.mathworks.com/matlabcentral/answers/994-wish-list-for-matlab-answer-sections#answer_1953
Editors of the forum should be able to accept answers by other contributors on behalf of the OP.
This ability should only be used whenever occasional users receive valid answers and never accept any of them.
Combined with a much needed improved search bar it will dramatically reduce the recurring of obvious and not-so-interesting questions.
We implemented this in a release on June 27, 2012.
The question must be at least seven days old before a contributor other than the asker may accept an answer.
If 3 or 8 editors vote frequently, not to say aggressively, the missing checkmark for acceptance would not be such a drawback anymore. See "Are we voting enough?".
Sufficiently long text will be truncated, both when posted in a Question and when posted in a Comment. No warning was given: the text just doesn't get recorded.
I did not trace the exact limit. Approximately the last 250 lines of a 1400 line file were discarded. The file was a bit over 100 Kb long.
The behavior inside a Question was odd: after the key point, I could type some extra text and it would show up in the display, but when I went back to Edit the question, the addition was gone.
My wish: something to make it easy to link terms to Matlab documentation as someone is posting a question. I read Answers to learn things about Simulink that I don't know about. The user may ask "How can I use the Waterfall Scope block to do xx" but I want to know "what is the Waterfall scope". Most people don't add the links manually - they assume everyone knows about the Waterfall Scope. Perhaps upon typing Waterfall Scope, the term would be highlighted; if clicked by the poster, the documentation link would be added.
That's something that has already emerged, it would indeed be great and save me lots of time looking-copy-pasting links.
Mention contributor's specialization to categorize answers. i.e: accepted answers (related to fields of medicine, engineering or financing simulation) should be marked to show contributor's understanding in the field. It will help filter out relevant answers quickly.
The bad formatting of code that occurs if there is an empty line in the middle of code that is indented by more than two spaces....
Hmmm.... seems to be working now, even with the icons above the edit box. I had used those before and something wouldn't work right - it would just insert asterisks or underlines instead of actually doing the italicization or bolding. Yeah, sometimes it's tricky with spaces. When it doesn't do what I expected I often try putting a space in front of it, like when you do a left parentheses and then click the link icon and paste in the link - it doesn't make it a link unless there's a space between the ( and the <.
_*bolditalic'*_
Once you figure it out how to do it as you type, it's pretty straightforward. The one that gets me is: (|doc fprintf|) v. ( |doc fprintf|)
I'm still undecided as to whether to edit the question for format the code (as Walter often does) or take Jan's approach of nagging the user to do it themselves so that hopefully it will be correct in the future without editing or nagging. What's that axiom about the man and a fish? Give a man a fish and he'll eat for a day, teach a man to fish and he'll have to buy a $10,000 bass boat, $100 reel, $200 worth of tackle and sit around and drink beer all day, or something like that. ;-) But I agree that this formatting stuff can be confusing and inconsistent and could be made easier and more intuitive and explained better. For example, have you ever tried to have both italics and bold on the same word?
@Jan, that sounds like you are trying to organize a strike.
@Walter: When I multiply the keyboard latency on your computer with the number of characters you have posted and take into account that a day has 86400 seconds only, I think I can proove, that you are specialized in Matlab Answers.
I appreciate your work and I do not dare to ask, if you'd switch to another Matlab forum, if it has implemented all important suggestions posted in this thread. It would be too bold to speculate, if your intensive and voluntary participation in this forum let TMW think, that the quality of the interface is good enough to support a working forum.
While TMW hesitates to spend some minutes include a more obvious description of the formatting of code in the page for creating a question, the hard core users spend hours each month to edit other ones questions and explaining the formatting in a comment.
Agreed on Walter and Jan's initial comments, if questions are about 'how to do it in MATLAB'. I am happy to accept those answers which solves implementation problems. But if the question posted itself is unclear, volunteers with a background to the problem can grasp a better understanding of what the query is - just a thought.
@Walter: and no.. I definitely cannot narrow down to your specialization :)
Currently Answers is averaging about 1 1/3 Answers per Question, with about 1/3 of Questions getting no Answer at all. That works out to an average of 2 Answers for Questions that get answered.
An "index exceeds matrix dimensions error" is still an "index exceeds matrix dimensions error" if it's in computational biology, finance, optim or "please post details".
Most questions only get one or two answers. It seems silly to devise a method to quickly filter a list of two items. Further, the voting mechanism should help with finding good answers (if people actually voted).
It's me who is concentrated on "please format your code" and "post any details". Does this belong to medicine, engineering or financing simulation?
Most of the problems, Ithink 99%, concern the implementation in Matlab, where it does not matter if an array is the solution of a PDE, a scanned image, a sound spectrum or random. Sticking to the meaning of the values usually blocks the creativity of the programmer, because only an efficient abstraction of the data allows for an efficient solution of the problem.
I remember some indexing problems, which could be solved by image processing tools like bwlabel().
People could miss out on a lot of good answers if that were followed.
If you look at the list of the 15 most prolific volunteers, only 3 of them concentrate on subject specialties, and the other dozen answer on a wide variety of topics.
And good luck defining my specialization: no-one else has ever been able to ;-)
As an editor I would like to have to go through at least one extra hoop before deleting something. Some combination of keys pressed (not sure which ones) and the mouse can cause actions you take to not even have a confirmation dialog before you accidentally delete something. This is dangerous.
Sean, you stumbled upon the MS Windows "Click Harder!" mouse modifier key sequence!
Be careful, as the sequence is confusingly close to the Halt And Catch Fire modifier sequence...
http://en.wikipedia.org/wiki/Halt_and_Catch_Fire
I accidentally deleted one of Oleg's posts while trying to upvote it. I am pretty sure my left hand was on one of the keyboard keys (alt, windows, ctrl etc.). But then all of a sudden the post was gone, the page refreshed, and a stream of expletives could be heard.
I have not observed this myself, except that there is no confirmation to delete tags.
Given how "serious" deleting things is and how infrequently people want to delete things, it is currently way to easy to delete things. I think an extra hoop (or two) for everyone would be warranted.
Further to not being able to turn off the per-keystroke updating of the preview (which is a big big pain when response times happen to be slow):
If we are going to have to live with keystroke-by-keystroke interpretation of everything we type, could we at least get some more advanced keystroke-activated editing functions?
For example, the equivalent of emac's ^T to swap two characters and position after the second of the two. This would save 3 keystrokes -- "backspace" to remove the first of the pair, right-arrow to position after the second, re-type the character that was previously first.
Saving 3 keystrokes might not sound like much, but when response times get slow, this can literally save more than 5 seconds (and another five minutes while preview catches up.)
Did anyone happen to mention that while preview is doing its hunt-and-peck updating, Firefox does not allow access to any other tab, so you cannot switch browser windows to do something productive while your brain cells die off?
Of couse we can copy&paste a dummy text at first to get over the magic limit, type the answer and delete the dummy finally. But it would be much easier, nicer and friendly to the public resources to change the limit to 1 character or add a button to disable the live preview if wanted.
I'm very disappointed that this trivial improvement is not implemented for such a long time now. The files@mathworks team is very polite, interested and active whenever I have contacted them. There seems to be another instance at TMW, who prefers to block improvements of the usability of this forum. The leaks in the usability have very negative effects to the advertising appeal of this forum, and I cannot imagine, that the responsible person does realize this enough.
Lukily I do not have to participate in this forum, when 1. the frustration level caused by the bad interface design gets to high and 2. when I have to post "please format your code" and "please read <how to ask a good question>" more frequently than real answers.
I succeeded with my HDF5 question - with a bit of copy&paste. However, there is a threshold, which The Mathwork could easily change. That question is less than 10000 characters including a lot of trailing spaces according to notepad++.
Per, it can take remarkably long to get to the point where the preview is turned off. 13600 characters in the test I just did. At anything more than 1/3 of a second per character, that would take at least one hour.
With long texts the automatic update is turned off and a [Refresh] button turns up to the right of [Preview].
At work this afternoon, at times the preview was 2 lines behind; last I heard our connection is a 4 Gb/s darkfibre that feeds directly into the 10 Gb/s research network in Canada that in turn feeds into the high speed research network in the USA which would probably exit over to the regular internet in Boston.
At home (wired, Cable network), I usually stop typing when the preview gets about 5 lines behind. Sometimes the delay gets bad enough that I start to wonder whether my computer is hung (seriously.) I've had time to get up, go down to the kitchen and pour myself some milk and bring it back upstairs before the preview caught up.
I use 3 systems:
1. 1.5GHz P-M, Win2K, 11g wireless Lan: near to unusable, > 1 sec per keystroke
2. 2.3GHz Core2Duo, Win7, 11g wireless Lan: fluently
3. 3.4GHz P4, WinXP, wired Lan at work: lame, 0.1 to 0.2 sec per keystroke
The problem is known at least since 13 months, see: http://www.mathworks.com/matlabcentral/answers/5560-do-you-suffer-from-keyboard-latency-in-this-forum . The keyboard latency is no proof of:
http://www.mathworks.com/ : "The MathWorks, Accelerating the pace of engineering and science"
@Walter: A virtual +1, but when I'd vote this, it would be harder to find in this long list of answers.
I have never observed this keyboard delay like others mention. Maybe my computer is fast enough, or I'm close enough to Natick MA, or who knows?
That before/when the magic editor powers kick in (edit and delete buttons showing up everywhere) that an email is sent to the editor with a minimum of the link to Guidelines for Editors and possibly additional guidance. Have the editors had any input into the guidelines? Is there any other guidance?
I had some input into them. The guidelines are not set in stone; the team considers feedback on them.
Some of the email amongst the editors has resulted in clarifications and interpretations. I don't think those are officially archived anywhere (but I probably have copies saved somewhere.)
The ability to mark up error messages in a question or answer so that they indent and behave like m-code but are colored red.
Fangjung, Matt Fig, and Paulo are already on strike, and Daniel is on strike on days that have 'u' in their English names.
If a few of you go on strike for a week or two we and maybe even TMW will notice!
Only a couple of the editors have noticed my end-of-year break ;-)
As there is no standardized format for error message (ah, for the days of VMS!), I don't think there can be automatic recognition. And people tend to copy only part of the message anyhow. I don't know if more than a small fraction of people would bother to deliberately format the error message. And that means that it would be another thing left to us. :(
Good idea, Sean. Currently I'm suffering from a dozen of unformatted codes and completely omitted error messages every day. I do not dare to dream of formatted error messages.
I assume, I need some holidays.
Now I learned the hard way how to loose the text I'm writing in Answer. The recipe is as follows:
- write text
- include link to a web site
- click [refresh] next to Preview
- click the link to test it
- text is gone forever - fool proof:)
At some occasion I have clicked [Submit] to save the text and clicked [Edit] to continue writing. That caused confusion. I received answers within seconds complaining that my question was hard to understand.
I want a [Backup] button next to the [Submit]/[Save] button. And a question "Do you really want to submit?" under [Submit]/[Save].
Or maybe better # a warning that if I leave the page text will be lost # alternatives [Proceed] and [Cancel]
--- Continue 2012-05-07 ---
Now I found another recipe. Click [Edit] while editing and the current text is gone. Yes, I said [Edit]. Please don't ask why I would want to do that!
A warning when the user is about to leave a page with "unsaved work" is commonplace. I think TMW should provide that! Sometimes, I get a bit upset with TMW giving us primitive toys.
@Image Analyst, thanks for the tip. I need a tool like that.
@Walter, I'll try to remember to start every Answer-typing-session by adding "-- To be continued --" at the end, do [Submit]&[Edit] to secure my text, and remove the line when I'm finished.
@Walter, yes you are right. However, I do mistakes all the time and wouldn't it be easy for TMW to provide a little safety net? And I'm not the only one!
Most of the buttons and highlighted text on the page invoke URLs, thus navigating you off of the current page that you had entered the text into.
Show / Hide comments is one of the few things done on the current page without disturbing it.
When you are planning to continue writing, you can put a written line, "To be continued".
To avoid losing your answer, if you are using Firefox I recommend using the Lazarus Forms Recovery add-on. It has saved me a number of times, including some cases where my complete system had become unresponsive.
Try Lazarus: http://www.pcworld.com/article/227948/lazarus_addon_recovers_lost_form_data_in_firefox.html
Works with Chrome also.
Another thing that would be awesome would be the ability to submit a published m-file as an answer.
Often I find myself answering questions with lines of code interspersed with lines of text. If that could include the markup and figures and output that would be great! And that way rather than someone having to copy and paste out the lines of code, they could just grab the m-file
Lower the reputation required to obtain some kind of 'editor' status. At least to the point where obvious contributors can fix stupid code formatting problems.
I don't know what reputation is required right now, but I haven't hit it yet. That's fine, but it's a little frustrating having to go back and forward asking someone to format their question when editors are either not around or not interested.
We greatly expanded how privileges are assigned in a release on June 27, 2012.
You can see your own privileges on your profile page. Heres a shortcut: http://www.mathworks.com/matlabcentral/answers/contributors/me
About two months per 1500 for me. My long term average is about 25 rep per day.
Lowering the boundary to (say) 1000 would make no actual difference at the moment with the current rankings: the only person nominally added to the list would be Matt Tearle, who already works for Mathworks and could have access that way if he wished. He has not been very active lately, though. All the people above 1500 rep at present are above 500 answers and 200 accepted, and no-one below 1500 rep is above either of those.
Short form: If we vote more and people will get to 1500 faster.
Long form: While I think tiered editor status might be reasonable (e.g., one level to add tags, another to remove tags, a third to edit/format posts, and a final level of being able to delete posts) this would require intervention from TMW (i.e., not likely to happen any time soon).
I also think 1500 is probably too high of a bar, again getting TMW to lower it is probably difficult. Right now 1500 rep requires about 600 answers/225 accepted. For all but Walter, that is a year time commitment. If we voted more, I would think one could get 1500 rep in 300 good answers, if not less.
It's 1500 - a pretty substantial time investment
Each question on the Home page should have an icon indicating whether there has been activity on the question since you last looked at it. That way you could see at a glance whether its worth reopening any particular question, rather than having to check.
The June 28, 2012 changes improved the situation, in that the index now shows the latest activity time and the nature of that activity.
Ouch! I wasn't involved back then
The time since last activity helps a _lot_ with that. We didn't have that until about October.
It would be easier to edit tags on Questions if the page did not get re-fetched for every single tag deleted. Especially on the questions that are longer, as getting back to the tag list can require scrolling down.
For example if there was an "edit tags" that pushed all the existing tags into a box in the format as-if they had just be typed in but not yet submitted (tag red-dot, tag red-dot, etc.) and that list could be edited (delete some, add more), and then when Save was pushed, the entire current list was pushed over to the database.
e.g., I am doing some work on removing the tag "help", and cleaning up the other tags for each affected message. There were about 175 messages with the "help" tag when I started, and typically I end up deleting at least 2 tags for each, and deleting 5-ish tags is not uncommon. That adds up to a lot of page fetches.
We implemented this in a release on June 27, 2012.
... this year.
I wonder if that mean we will eventually see that implemented ...
Excellent suggestion Walter. Thanks for taking the time to clean up the tags.
Larger font for the Question and comments on the Question, please. The Answers and comments on Answers appear to be a point or two larger, and it does make a difference to my aging eyes.
Sometimes I have to Edit a question just to be able to read it clearly :(
We bumped up the default size for questions/answers/comments on June 26, 2012.
Only one + is needed
WOW!! ... THANKS!
ctrl + (zoom in)
and
ctrl - (zoom out)
Greg
Or hold down the control key and move the mouse wheel, which works in most apps.
Could you just zoom the browser? e.g. ctrl + + to zoom in on ie8.
In lists of Questions, it would be convenient if there was an annotation to indicate that you had Answered the question; also convenient would be an annotation to indicate that you had Commented somewhere in the Question (including on an Answer.)
The annotation does not need to be much: it could just be something like a colored asterisk, or an '(A)', possibly near the number-of-answers box.
This would ease the load on my aging short-term memory.
- I have not had time to investigate to pin down circumstances, but it appears that now when a search query of multiple words is given, the searching is no longer always an AND search. If two words are searched on, and one of the terms occurs "relatively few" times (or not at all) compared to the other, then the search returns all the Questions in which the more common term occurred, ignoring the less common (or no-match) term.
- There is no documented way to indicate that a multi-term search is to be AND, OR, or PHRASE
- The order in which matches are returned can be odd when multiple terms are given for the search. Sometimes about a 10 Question matches are shown that relate solely to the first term, then 20-ish Question matches are shown that relate solely to the second term, then the remainder are shown by age -- resulting in two age-related jumps in the list.
+1. I did not vote this before, because the search bar of Answers is such weak, that I do not use it. But actually a powerful search is a fundamental feature of database of questions and answers. We have 27'273 threads currently. All messages have tags and time-stamps. It should be easy to find a message based on the information like: Walter has written something about periodic functions about 4 month ago.
using the keyword "and" between phrases does appear to change the results to be more AND-like.
- If all entered tags for a post are the exactly the same, then the tag is displayed with a count after it, but the construct only counts as a single tag for the purpose of the code that determines whether tags are editable or not. For example, if a posting was tagged with matlab, matlab then it would show up as matlab(2) but you would not be able to edit it to remove the extra copies of the tag. Workaround: add any other tag so that all tags become editable, do the pruning, and remove the temporary tag.
- observation: In any list of questions, at most 5 tags per posting are shown. The limitation is 5 tags, not the number of characters occupied.
- observation: The oldest one-thousand nine-hundred and ninety-nine (1999) postings are all marked as having been updated Feb 24, 2011 (or sooner). I guess there must have been some kind of software update then.
- I mis-spelt a commented_by: query as comment_by: . It turns out that comment_by: followed by anything returns the entire list of postings (24000+ of them) instead of returning an error. For example, comment_by:foo
The following is not formatted correctly - note that there is an amply line in the middle:
if n < b
n = n + 1;
b = b + 1;
end
To display the real contents spaces are replaced by #:
##if n < b
####n = n + 1;
####
####b = b + 1;
##end
Replacing the empty line by '##' or '' does not help, but '#### ' keep the correct indentation. Using more than 2 leading spaces for all lines solves the issue also:
###if n < b
#####n = n + 1;
#####
#####b = b + 1;
###end
is displayed as:
if n < b
n = n + 1;
b = b + 1;
end
but I assume then it is preformatted text and not code anymore.
Formatting the code seems to be a hard problem for newcomers in this foum. It is very important, that this key feature of MATLAB Answers works reliably. In opposite to WYSIWYG there is still a certain level of Guess What Is Needed To Let It Look Like You Want It.
I still think this behavior is undesirable. It breaks the model we have been given for how indentation works, and replaces it with an undocumented and non-intuitive model.
I received this explanation from the support team:
"I heard from the MATLAB Markup engineering team that this is the expected behavior. We do not have any way of including blank lines in the same block in Publish right now. Every blank new line ends the previous block, and starts a new block."
Thanks to the TMW team for the exact answer.
I think this is *not* a useful style for formatting.
Since the maintenance a few days ago, the site has been slower than before.
At home, it can take an undetermined length of time to fetch the user avatars unless they already happen to be in local cache. At least that was the situation about 7 hours ago; I do not see the problem at work at this time (My IP at work hits a different edgesuite server than home does.)
Editing of existing content degraded over time after the maintenance, affecting more and more questions, and now affecting all questions. The symptom is that instead of getting the edit window with the buttons for bullets and Code and so on, I get the bare text small edit box. This does not apply to writing new questions or to writing new answers, but affects editing anything existing (e.g., to reformat an existing post)
I am using firefox at both home and work, but different versions and different OS; I am not observing problems on other sites.
It appears that reordering of questions is no longer done immediately, in the sense that now if I update a question and then ask for the list of questions, the newly updated question will not go to the top of the list for a period of (estimating) 15 to 20 seconds. The page that is shown will know the proper update time for the question, if it happens to be one of the ones visible on that page: only the order is affected. We could probably get used to this delay; the confusion from it is more from expecting instant reordering as we had before.
Turns out that my browser was wonky because it had a half-installed update of NoScript waiting. It was fine after a restart.
Walter and I have followed up offline regarding the avatars. If anyone else is having an issue with avatars rendering, please feel free to contact me or files@mathworks.com.
The issue surrounding the content edit box is resolved as of yesterday.
Thanks for the follow-up clarification!
The custom avatars are responding promptly; the default avatar never gets fetched, not even after 6 hours.
Hi Walter, thanks for reporting these issues! We're looking into why the avatars are rendering slowly. We're aware and investigating the issue of editing content as well as reordering of questions. We'll follow-up on these.
It looks like it might just be a problem with the default silhouette avatar used for people who have not set their own.
Confirmed, at home I am still having the problem with it being unable to download the avatars.
The status of questions should be updated promptly when they are answered. A search with "Unanswered" selected returns several questions with one answer.
Excellent! Thank you.
Happy to say that this issue has been resolved (as of 1/26/12).
Hi Andrew, thanks for pointing this out. We'll have a fix implemented this week.
Is it just my imagination, or did the site switch from displaying 50 Questions per page to 100 Questions per page overnight ?
I also notice today that when I tell my browser to open a Question in a new tab that the tab name given for it is "MATLAB Answers - MATLAB Central" . I thought the tab names were the first few words of the question title, but now I am not certain. It appears the HTML Meta Title property is used for the tab name; as I often have more than half a dozen Question open at a time, meaningful tab titles are useful.
Good news - both of these issues have been resolved. Again, thanks for reporting them.
Whew! It was like looking at someone and wondering whether they just got their hair cut or if you simply hadn't paid attention before!
Hi Walter, this was not your imagination at all! Both of the changes you noted were unintentional and the behavior will be reverted back.
As always, thanks for catching this!
Lindsay
I would like to propose an "achievements" system similar to existing one for other Q&A sites (for example stackoverflow). I think it might achieve a bigger participation and interest of the community users.
You could be right about it achieving bigger participation; my concern at the moment would be with the programming effort, as it already takes a number of months to get fundamental feature changes through.
Please add a latest activity line in the listing of topics.
It is very annoying to have to open the topic just to see if I was the last responder.
Hope this helps.
Greg
I agree, and voted for it. I thought it might already be on the list somewhere but the list has gotten so long that I don't feel like scanning 113 (literally!) answers and comments to check anymore.
iMatStore
Instead of paying for a complete toolbox, you could go to the iMatStore and download a single track. Who listens to a complete toolbox nowadays?
:)
Being able to accept more than one answer. Just posted a question regarding dynamic array names. One Answer, answered the question directly, is correct works and I will use in the future. Another answer however made me look at my problem slightly differently, provided an elegant solution that will be easier to implement than my original Q? Although both are correct answers and both will be used I can only give my Answered calmer points to one person.
I found a formatting problem.
w
y
%z
%p
q
The formatted version of this shows the two comments indented one level, but in the source, the comments are at the same indentation level as the "y" lines. The source has the "q" line at the same level of indentation as the "w" line, but they show up at different indentations, and the "q" line shows up at the same indentation as the "y" even though they are different indentations in the source.
The problem does not happen if you remove the empty line between the "y" and the commented "z", and the problem does not happen if you remove the "w" line.
Do I win this week's, "Now, that's obscure!" award? ;-)
This is a copy of the source, with each space replaced by a plus:
++w
++++y
++++%z
++++%p
++q
So, I tracked this down with the team responsible for the markup. It turns out that the formatting you're seeing is in fact the correct result.
The first two lines are indented with two spaces, making them part of a code block; two spaces get subtracted from the result in each line.
Then there is a newline, which ends the code block.
The next line starts with 4 spaces, making it a pre-formatted block, not a code block. And the white spaces at the start of each line are preserved as in the source.
If, in MATLAB, you paste your source into a script and publish it as HTML, you should see the same formatting.
The answer was there all the time, it took a small accident to make it happen... AN ACCIDENT
Thanks Walter, you certainly take the week's award. Interestingly, if you indent with one space the comments line up properly. We'll take a look!
Some tags have become non-functional. See http://www.mathworks.com/matlabcentral/answers/8026-best-way-s-to-master-matlab and click on the tags there
learning-matlab
or
studying-matlab
Both of them say there is no search result.
These two tags are, as best I know, the most recent two tags to be added to that Question; I added them late last week (before the software upgrade.)
Thank you!
Walter - These tags have been fixed.
Thanks for letting us know.
Helen
The new interface and all seems nice. It looks like TMW is trying to address some of our requests. It seems odd to me that the "Latest activity" doesn't register comments or accepted answers as being activities.
My wish is that the "Latest activity" was really the last activity that caused the question to jump to the top of the history. Even better would be for Latest activity to provide a chronicle of all activity and links to the relevant answers and comments.
As of June 27, 2012 the following count as latest activity for a question:
- Edits to the question
- New answers
- New comments on the question or an answer
- Closing or reopening the question
Comments against the original Question show up as activity, but comments against Answers do not, as best I can tell. Likewise, edits against the original Question show up as activity, but not edits against Answers.
Or maybe not. Seems like sometimes comments show up as an activity and sometimes not ...
It looks like latest activity now handles answers, comments and edits.
There is an obscure error that I am trying to chase down that has to do with the space-space formatting.
Sometimes if a user has a series of lines beginning with >>, the lines may look proper in the editor, with the leading space-space seeming to be there for each line, and the Preview might look fine -- and yet the finished display might be wrong as if they were not in a formatted area. For example,
>>hello
>>there
might show up as
>>hello >>there
in the finished posting.
I had a chance this evening to cross-check an instance: in that instance, the paragraph before did end correctly (no space at the end of it for example) and there was a completely empty line before the series of >> lines, and each >> line appeared to start with two spaces, and the Preview was fine in the editor, but the composted version was wrong. (Another editor happened to fix the formatting while I was in the middle of checking things out. Oh well.)
The first time I encountered this situation, I went through every line and manually deleted what looked like two leading spaces (just in case they were something else) and typed in spaces myself, and saved, but that did not make any difference: I ended up re-typing the lines, and that worked just fine.
None of my hypotheses about the cause have born out yet; my tests have all come up negative, unable to reproduce.
If you are an Editor, and you happen to notice a situation like this, where a block of code looked unformatted in the Answer display, and yet appears properly formatted in the Editor, please leave the posting uneditted and report it for investigation (I'm presuming here that Mathworks can examine the binary format of what is really there.)
The sort of straws that I am left clutching at are possibilities such as "Maybe the poster copied some UTF-16 there..."
I believe I have tracked this problem down.
In the cases where a Question or Answer looks like it has not been formatted in the rendered display, but the source for the Question or Answer looks like it is formatted, then the cause is usually that someone else edited the Question or Answer between the time you displayed the Question and the time you went into the editor. The rendered display reflects what was fetched when the Question page was fetched, but the editor reflects what was stored in the database as of the time you entered the editor, rather than being an editor of the content as of the time the page was fetched.
I have not observed this issue for a number of weeks. I did observe it a few times after the above report, but each time someone happened to edit the posting before I could track down anything concrete.
Walter - We are unable to reproduce or to find posts that this may have happened to. If you remember which posts you saw this on we can test further.
Users are required to enter tags, and a recent change made it impossible to delete the last remaining tag -- but ever since the tag cloud was removed, there is no explanation of what tags are, no promotion of the fact that they are available to find Answers, and no user-obvious way to search by tag. And no means of finding out what tags are available or of searching to find keywords within tags.
The tag clouds were a bit clumsy when the space to paint them was variable, but the clouds did give the users a feedback of the importance and usage of tagging.
I'm not really convinced that automatic clouds are the way to go, but the refine-by portion of the left side of the index page should link to a tag search facility.
I agree. And some tags are very useful, like tag:tutorial.
When using search, you can sort by date answered/asked, and number of answers/votes, but you can't sort by how relevant the questions and answers are to your search (even a simple # of times each word occurs).
Filters based on product/category tags are good, but will only find questions that have actually been properly tagged.
Of course I can get round this by pointing a google search at Answers, but it'd be handy if the inbuilt search function had a go at it.
I still think that they should just add a Google search box. You don't have to do everything yourself.
I dont know someone have already pointed it or not, Currently 'comments' have only option to delete. It shold also be allowed to Edit the comments.
Yes, http://www.mathworks.com/matlabcentral/answers/994-wish-list-for-matlab-answer-sections#answer_1427 . It is the very first comment in the highest voted request (mark up comments); you may have to "expand comments" to see it.
I was sure, this has been asked already. But I cannot find it - perhaps it was another thread.
It would be even better than the current behaviour, if comments can be edited, but *not* deleted.
I wish TMW wasn't as liberal about disclosing "hidden" email addresses. If you to hide your "Community E-mail Address", but then use the web interface to contact someone, your email address gets sent on to the recipient. Even if you never contact anyone, if they contact you through the web interface and something goes wrong in the email delivery (or you have a vacation message set), the returned email is sent directly to sender with your email address revealed. Doesn't seem so hidden to me.
I originally asked this question in:
To save, to mark or put on a favorite folder answers (from other peoples) that you think that is interesting to you.
You could just make a folder for those bookmarks (of those special Answers you want to save) in your bookmark panel of your web browser.
That sounds like the same thing Andrew suggested in
http://www.mathworks.com/matlabcentral/answers/994-wish-list-for-matlab-answer-sections#answer_4972
As of last night (September 21, 2011), it is not possible to delete the last tag of a Question leaving no tags at all.
This can be a bit of a nuisance for editors, who may be removing inappropriate tags but do not necessarily have the time or subject knowledge to work out "good" tags for the Question.
It is now possible to delete all of the tags from a question. However, please try to replace inappropriate tags with at least one "good" tag. Here's our standard advice on adding and deleting tags:
- Use correct product names in all references
- Add standard-usage tags as needed to facilitate searches. For example, use the format r2010b for release numbers (not matlab r2010b or r2010b 7.1). The goal is to avoid diluting the usefulness of tags by creating variations; for example, rand, random, randomness, and randomize.
- Delete tags that are meaningless or obstruct searches.
- Delete products that are not related to the question.Separate tags lumped together by mistake: for example, delete matlab fft and add matlab and fft.
A man, a plan, a canal -- panama!
Add a "bad_tag" tag. (A sentence with only a's as vocals!)
OMG - they finally ordered "my answers" newest to oldest. (of course I had already blindly clicked to the last page)
Yup. There would have been a two minute window during which your response was 210 minutes old or older (3.5 hours rounds to 4) but my response was less than 210 minutes old (2.5 hours up to but not including 3.5 hours, rounds to 3 hours)
The date of the update in question was Tuesday June 28, 2011, which was 12 weeks and 1 day before the Wednesday September 22, 2011 update. Tuesday of next week (September 26th) would have been 3 calendar months, so Monday 12th would be half a calendar month earlier (30 days hath September); by that reasoning, anything from July 12th to August 11th would round to "2 months" and before that would round to 3 months. Or if you work on the basis of a "month" means 4 weeks exactly, then anything 6 to 10 weeks before September 22nd would round to "2 months", which would back-date us to July 13th if my count is correct.
This presumes the rounding is the traditional [-1/2 to 1/2); we don't know that is the case, particularly since we do not know whether it goes by weeks or calendar months.
For an example, look at http://www.mathworks.com/matlabcentral/answers/16475-how-to-assign-zero-pixel-location-of-one-image-to-other. Walter's comment to my first answer originally was "12 minutes ago" and my second answer was just two minutes earlier at "14 minutes ago." Yet now when I look at them Walter's is a whole hour later than mine ( 4 hours vs. my 3 hours) when actually his post was actually only 2 minutes later, not a whole hour later. I guess it just depends on where they put the dividing line when splitting up times into "friendly" times. I agree with Walter that friendly times are okay if the real, actual time stamps are there and easy to see . . . but they aren't there at all.
I want to hide the actual time stamp. Boss might find you spending too much time during work hour on MATLAB Answers! But, if it is true that there isn't necessarily more than a few hours difference between "2 months ago" and "3 months ago", then it's terrible. I need to make a wish-list, or maybe it's already there.
@Fangjun: proecsm's posting that we are replying to is the duplicate.
@Jan: I read the HTML source long enough to work out the pattern, and then worked out the short-cut. I described it in the comments to http://www.mathworks.com/matlabcentral/answers/994-wish-list-for-matlab-answer-sections#answer_1459
@Image Analyst: I don't mind "friendly time" as long as the real time stamps are there and easy to see.
@Walter, my point is that I didn't make a "duplicated" "wish". In fact, I didn't make that "wish" at all because the feature was already there. I was complaining that the feature was lost. The system update happened 2 month ago probably screwed it up.
Walter, that's why I'd just prefer they used the actual time date stamp (like most every other forum) rather than do the extra work of converting it to some "friendly" time.
@Walter: How do you obtain the anchor of an answer as "#answer_14668"? Are you reading the HTML-source?
The date approximation code does rounding on converting minutes to months. There isn't necessarily more than a few hours difference between "2 months ago" and "3 months ago".
The feature was there 2 month ago. It was ruined by the system update 2 month ago. Matt's wish was 3 month ago.
Duplicate "wish" -- covered by http://www.mathworks.com/matlabcentral/answers/994-wish-list-for-matlab-answer-sections#answer_14668
Yes! I noticed it too. I am waiting to see if any updates will make it to the top. That is the most important feature to me. I have been nagging about it for a while.
If you make two updates to the same Question without refreshing the page, then after the second update, the time of the first update will continue to indicate "less than one minute ago" until you refresh. That is, adding a new Answer or adding a new Comment does not trigger the refresh logic that updates the update times.
If you make an update to a Question and while you are composing your update, someone else updates the same Question, then until you refresh, you will only be shown their update if it happens to be an update to the same series of Comments (either to the base Question or to the same Answer); making an update does not trigger the refresh logic that looks for changed contents (edits) or for new Answers or new Comments on other answers.
These are both bugs, with the first of them being fairly low priority and the second sometimes having the potential for confusion.
Ran in to another instance of this tonight. After I submitted a comment, my display showed that the Question had been submitted about 32 minutes ago, that Oleg had Commented about an hour ago, and that I had Commented less than a minute ago. That is, when my Comment was received, the list of comments was fetched from the database and Oleg's comment of about an hour before was added to the list before my comment, but the time marker for the Question itself was not refreshed when my Comment was received.
It is convenient forthe readers to insert a link to the documentation, when a specific command is suggested. Getting this link is less convenient. Fortunately Google lists the TMW documentation whenever I ask it for "Matlab <command>" (replace <command> e.g. by "plot" without quotes). Then I copy this link and insert it in my replies.
A new tag would be easier, e.g. $command$, which is converted to <http://www.mathworks.de/help/techdoc/ref/command.html COMMAND> dynamically.
- This is faster to type.
- The dynamical creation at displaying the pages avoids dead links, if a command is moved from a specific toolbox to plain MATLAB.
- The more frequent the readers of this forum are forwarded to the documentation, the more accustomed they will be to use the docs voluntarily. And this will reduce the work for the technical support.
Btw: Thanks to TMW to care for appearing the documentation on top of Googles answers!
This is in some ways related to my earlier idea about easy links to specific answers, the Answer that begins "It appears that answer numbers and comment numbers are unique."
Totally agree. I use the link to my tutorial which has on top many link and mong those to the documentation, then I type the name of the command.
Although I think, that nobody will be able to find this message:
The margin on top of a section should be smaller or equal than below a section. Currently it is large and it is harder to see, what belongs together:
Typed text:
<newline>
*But the integer saturation does not cause an error message:*
<newline>
<space,space>for i = uint8(1):uint8(1e4) % !!! uint8(1e4) is 255!
<newline>
*Creating the index vector explicitly is slower:*
<newline>
<space,space>v = 1:1e4;
When displayed in the browser (Firefox 6.0, Win7), the vertical space under the bold lines (thought as header) are larger than the vertical space under the grey source:
In addition the quality of PNGs is reduced by upscaling by about 20% and perhaps a lowpass filtering. Showing pixel images in their original size would be nicer.
I just check on Firefox 3.6 (on Linux Ubuntu 10.04), and the image _does_ show a bunch of artifacts.
I do see the differences in spacing and I cannot think of a good reason to retain it.
The image is already full size for me, at least in TenFourFox (a Mac PPC port of FireFox 4)
@Walter: I've tried to get the text a little bit clearer. The image is shopwn in its original size, when I open it through the context menu->view image. But then the charm of embedded images is lost.
Are you referring to the spacing between lines?
Note: in my browser the quality of the PNG is exactly the same as when I visit the image directly.
Coming later: the implementation.
- The inverse of being able to mark a question to follow: being able to mark a question (or even an author) not to follow so that it doesn't show up. (Obviously if I blocked an author, I could see that person's answers, but not their questions. Maybe this is cold-spirited but it would be nice for the doit4me_Daily crowd)
I quickly traversed the answers but guess I missed that (red_faced_emoticon!)
I do feel like utilizing that would be in poor taste, but it really would be nice some times. As far as not watching threads, the energizer bunny thread from a while ago comes to mind. Or if you're not interested in meta stuff but just answering questions, this thread could be seen as annoying along with all of the other Answers related ones.
An interesting variation on an idea I wrote about above, "And possibly some respondents may wish to "subscribe" to being warned or blocked from replying to someone found to be unproductive to deal with."
Free MATLAB swag for frequent contributors. While striving to get added to http://www.mathworks.com/matlabcentral/answers/1058-contributors-metainfo-reputation-and-more might be enough motivation, I think free stuff would be more motivating.
Maybe they could start small with a MATLAB answers sticker (not just the standard MATLAB sticker, but a customized one to reflect the contributions to answers). then as you contribute more you could get a mug, tee-shirt, USB key, etc.
In a previous life, the expression used was "I've been waltered"! (Note: this was not related at all to the much later use of the expression in a tractor forum.)
@Walter: And for the majority of users a "WalterdidIt4me" collection.
I've started a line of "doit4me" clothing ;-)
A wish to be able to access a full list of the active Tags (even the ones only used once or twice.) Required to find and investigate unusual tags without going through all of the postings.
Many questions get asked again and again. Why? Because people are not searching for previous answers before asking questions.
Stackoverflow suggests some similar questions that have already been asked when people are typing in their questions. It's therefore easy for them to see that the questions already exist.
This system also relies on good questions being voted up, so they appear at the top of the list. I have asked several (dare I say) well-worded questions (and usually answered them well too), with a view to providing a resource for people to refer to. This effort is wasted by the questions and answers receiving no votes, and the fact that people never search for them anyway.
So, TMW please implement question suggestions like Stackoverflow, and regular users please vote for well-worded questions.
Just curious, why is Matlab answers separate from Stackoverflow? I would like to combine all my reputation points (low as they are) in one place.
I would like to have a way to see when someone's account was created and how many questions that person deleted (maybe even the names of the questions deleted for those who are editors), this is because I clearly remember having answered several questions from someone but all those questions were deleted and the very same person posted a new question yesterday, I replied and that person exhibited the same old behaviour that triggered my memory, he/she didn't even look at the links I provided and asked something that isn't MATLAB related, I didn't answer the specific question because I would waste my time, this is MATLAB Answers not Control Engineering Answers.
We implemented "no delete if already answered" in a release on June 27, 2012. Click on the "Link" button next to an answer or above a comment.
Although it is true that in theory this forum is about MATLAB, I find that in practice there are a lot of more questions about theory posed, or questions about theory that are disguised as MATLAB questions.
A fair bit of the time, "How do I do {something} in MATLAB" really means "I barely know what {something} is, but I have to implement it in MATLAB: how do I do that, preferably without ever having to learn anything about what the {something} really means or the mathematics behind it?"
I do admit that is the cynical view from having dealt with large number of students; I _do_ see people (including some of the students) who are interested in their topic, but are lacking the background in it and are also perhaps weak on their research skills (or research initiative) to investigate before trying to implement.
Or at least answered by someone else. If you want to delete a question you answered yourself, I don't really see an issue.
Agree on: no delete if already answered.
And not delete someone's really thought out answer. (I know I like to reference old answers frequently. I also keep all working code I post here (and images) in 'ans###.m' script files with ### being the month and day (yes I'll face a mini Y2K next January :) ) )
I think it would be fair to make it so that an answer couldn't be deleted by its OP once someone had answered. I guess they could still edit away all of the information but it would at least appear as a question on their profile.
How do you guys find the new comments or answers in this big mess without having to inspect all of it? So darn frustrating!
Yup!
+1 Oleg! Exactly what I do.
nice tip Oleg, doing the same with Opera from now on :) thanks
On Mozilla: ctrl+f and I look for minutes or hour
NOT having my answers be listed oldest to newest. Having to trudge through 14 pages to see my recent answers is no fun...
Perhaps if we could choose the order from a drop-down, this would be more functional.
2011 09 21 - sorting order is now most recently modified to least recently modified for "My Answers" button.
Searches are case-sensitive.
For example, searching for alphadata returns a different set of postings than searching for AlphaData
This is a bug, at least in the absence of a clearly indicated (and easy) method to make the search case insensitive.
When you delete a post you are reading, often the attempt to return you to the main MATLAB Answers page results in an error about the page not existing, and one is usually unable to get to the main page for 15 or more seconds.
(This bug was previously reported in email and is/was being worked on, but I realized I hadn't documented it here.)
When you delete the post you are reading, you are returned to the main MATLAB Answers site instead of being returned to where you were in the list (e.g. you might have been on page 2 or 3, reading something past the most recent 40 threads.)
A "Views" counter for the questions.
A "How many users are online" string somewhere on the home page.
Oh, o.k. You confused me with your technical speak. ;-)
"logged on to MATLAB centeral" is the same thing as "authenticated to the web site" that I wrote about.
How about how many people are logged on to MATLAB central? That would at least give an approximation...
Perhaps a bit difficult to define. If I'm composing a reply, in at least one sense I am "online", but the composition is taking place entirely in my browser.
Counting the number of people authenticated on the web site is not useful either: I'm typically authenticated for over a week at a time.
"Transactions per minute per distinct IP address" would start to get closer, but that would be skewed if multiple people are reaching it through the same IP address -- e.g., large portions of the organization I work for all interact through the same public IP address with Network Address Translation.
That would be cool!
I am adding this again because it is REALLY, VERY annoying!
We should be able to put quotes around a paragraph or sentence and have it show up as such!
Right now if you write a big long response of several paragraphs and code, then put one little sentence at the end on its own line and in quotes, everything you wrote will be wiped out and only the line in quotes will show up - and without the quotes. When you then click on Edit, hoping that somehow your work will have been saved and just not shown, you find it is gone...
This is utterly ridiculous! At least add a button so that we can highlight a section in the answer and put it in quotes...
Thanks Helen!
Matt and Walter - I have logged a bug report for this. We'll look into why this is happening.
I'm pretty sure the apostrophe / double-quote behavior is a processing bug rather than a deliberate design... but surely it could be fixed!
And yes, I did encounter this problem again the other day. Fortunately I had Lazarus installed and was able to recover the text that way and edit it.
I don't know if it is possible, but it would be nice if when one deleted something (e.g., a comment one made) that the new page could be positioned "near" where you were.
This should be possible for sure if one was deleting a Tag or Product.
Marvelous!
All of these deletes should happen without a page refresh now, so the browser should remain right where you were.
I was going to comment on something Walter brought up 7 days ago. It's up there about 25% of the way down this list. It starts "A way to get actual time-stamps for items..." (It was posted 7 days ago but for some reason, apparently a known bug, it's listed way up there in between two 4-month old postings.) But I don't know if my response would ever be found among the 83 other answers buried way up there in the middle of the list. So I decided to add an "answer" at the bottom of the list. At least that's my intent, probably just like Walter who may have been surprised that his post got buried way up there in some strange location between much older posts. Anyway. . . on to the suggestion:
When you go to Google, it does a half way reasonable job of determining when you last visited and it flags the new ones since you last visited as "New." I wish MATLAB Answers did the same. I also wish there were accurate time and day time stamps (like Google). If I come back to visit answers, I have no idea how far back I need to go in the list to see the new ones. But if I know that I last looked here at 2:00 p.m., and the posts were time stamped right there on the main screen, then I would know which ones were new. If they're not time stamped, I may waste time opening up threads that didn't have any new stuff at all since I last visited.
So then I double click on it (my nicely time/date stamped listing) to open it, and see 83 answers and dozens of comments. Like Walter said - where is the new one(s) located? It could be buried deep in pages of old stuff. There needs to be some way of indicating which ones are new to me, or at least recent (say within the last 24 hours or so). Maybe make them a different color or boldface or something.
Coming soon: the book.
Be able to mark a question and its answers into "My QA Collections", just like "My Questions" and "My Answers". This will provide an easy reference for any QA that I am interested.
Asking a question, sometimes I forgot to fill "keyword" field and press OK button, the error occured, but by browser (Mozilla) button "Return (Undo)" I have seen EMPTY form. That is, all text I have to retype :(
I would call this a bug.
Easier to change JavaScript then to install Lazarus :)
That looks like a great add-on!
If you are using FireFox, you can get your question back again with a few clicks if you install Lazarus Forms Recovery, https://addons.mozilla.org/en-us/firefox/addon/lazarus-form-recovery/
This is, technically, the same problem as I wrote about earlier in the note starting "When there is a problem posting" .
On the other hand, I don't think that it is acceptable behavior that you get the blanket error page if you do not fill in the keywords section. The Save button could be disabled until that field is entered or a javascript-level pop-up error could be generated instead of submitting the question and letting it disappear without a trace.
It would be nice to have more flexibility on date selection. For example, it would not be uncommon to want to look at everything that had been modified since the last time one checked; that could be a varying amount of time, but "1 day" would bracket it much better than 7 days; likewise a time frame corresponding to checking on Monday after not having checked since Friday would likely be common.
I did do some experimenting with supplying my own time frames, but only the time frames that are selectable from the index were recognized even (not a syntax matter; the syntax was clear.)
A way to get actual time-stamps for items instead of "about 3 months" and so on. For the purposes I am thinking of, having the time-stamps show up in the HTML but not visible would probably do (but an option to see them clearly would be nice.)
random reshuffling of Answers is a bug.
Quite true, Jan. I had to go looking carefully for new responses in order to see yours.
If I recall correctly, Randy already termed the random reshuffling of Answers to be a bug. The sorting by votes is not a bug, though, so if something newish gets voted up in a long list then it would be easy to miss.
And sort the answers according to the date on demand. E.g. it will be hard to find this answer after I've voted for it, because it will vanish in the heap of the other 77 answers. Sorry.
Being able to revoke an Accept "at least for a grace period" was suggested earlier.
I would suggest that Editors be able to revoke Accept. It is not uncommon for a user to Accept much too early. Sometimes too the Accept looked reasonable at the time, but the user then either finds the solution to be unusable in the situation, or has additional related questions that are easier to deal with in the existing Question than in a new Question.
When a long Tag is entered and the total Tags length is enough that one or more are positioned on additional lines, then the positioning of the first tag on the second line may be messed up, not starting until a few characters before the start of the long tag. This happens especially if there are multiple long tags.
Sometimes users enter entire questions as single tags, or try to separate tags with semi-colons or the like: long tags can result. And sometimes there might only be a single long tag that has a spelling error (along with shorter ones); when I go to fix the error by entering the corrected tag before deleting the old one, the display of the tags usually messes up.
Refreshing the page does not help: it is not a momentary glitch. It is possible, though, that it is browser-sensitive: I have seen it with Firefox 4 and TenFourFox
Some way to prevent questions being deleted without a good reason!
I'm getting really annoyed with people whose response to your help is to disappear and take with them both their question and your response to it. It's happened more than once to me, and not necessarily questions where the poster was doing something dumb -- perfectly normal question, valid answer(s), then... floop! It's gone.
We implemented a partial constraint on this behavior in a release on June 27, 2012. You can no longer delete your own question once someone has answered it.
What really upsets me about a question that I have answered being deleted is I never can remember who asked the question. My feeling is if you are going to delete a question that I have answered, then I am not going to answer any of your other questions (at least until I think you have learned better).
When one is adding a Tag, one can type a comma after the text to signal that the text is complete and the tag corresponding to the text is to be added.
When one is adding a Product, this is not available: one must click a product list entry to have it added.
It would be nice if one could also use the comma in the Product list. In particular, this is a nuisance when one is trying to select the generic product MATLAB (the most common Product) or the generic product SIMULINK (probably about the third most common Product).
You should be able to do this now (I can't recall when we added it).
Per-user (user-controllable) default tags.
For example, if the user could (was prompted to) set up tags corresponding to the MATLAB release and OS + Version that they are using, then it could save questions about "Which version are you using".
Further refinement: as users might not notice this facility, perhaps an ability for people (editors?) to trigger a "suggestion". I'm not sure how that would work... perhaps in the form of a URL sent to the user that they could click to confirm, or perhaps in the form of a URL sent to the user that they could click to take them to the configuration form.
These could also be update-able fields set in each user's profile when registering. Something like:
User Name: Matt Fig
Versions in use: 7.01, 2006a, 2007b
etc...
etc..
Then they wouldn't set the default tags.
What if they have multiple versions?
I have noticed that some users are interpreting the {} of the '{} Code' button as indicating that the proper way to mark a section as being code is to enclose it within a { } pair.
I'm not sure the {} part of the button label is adding any clarity to its purpose, and it is confusing some people.
I believe, that the "{}" is a relict from experiments with different markup methods. For the current method, there is no cennection to curly braces.
Bracket matching, please. Even if the first version doesn't account for quoted brackets, it would be a help. Restricting it to "Code" sections would be fine (and would make sense.) Best would be deep enough parsing to recognize problems per-statement on-the-fly, but even having to select an area and invoke the bracket-checking function on it would be good.
A tutorial section for posts like the following: http://www.mathworks.com/matlabcentral/answers/6200-how-to-ask-a-question
I like your approach to this. I had thought of creating a similar page, but the big problem is keeping the page visible for new users. A tutorial section might do that.
Wish: that the main page for Answers would stop saying that the page was not found.
(I got in directly through the Contributors page, and I cannot see any issues; the database still seems to be active, but the main page is fried.
This problem is now fixed! Yeh-Haw!
I have determined that this likely a problem with one of the first 8 most recently added messages (added before my series of tests, that is.)
Sometimes Questioners perceive negative value in some responses, and ask that the respondent not reply to them in future. Possibly those Questioners would prefer if there was a mechanism to actively block specific people from replying to them in future.
And possibly some respondents may wish to "subscribe" to being warned or blocked from replying to someone found to be unproductive to deal with.
I shall not defend the merits of these possibilities, just raise them because I happened to think of them. (Besides, the more I think about the second of them, the more seductive the idea becomes...)
Wish: an editorial function to move an Answer to be a Comment on a different answer (or even the question itself), while preserving appropriate ownership. With any Comments to the Answer to be moved as well to become additional Comments.
Questioners post their responses as new Answers relatively often; cleaning up after that by copy-and-past is tedious and error-prone and doesn't preserve ownership.
Bonus Wish: move a Comment to become an Answer, preserving ownership and allowing appropriate credit. Especially during back-and-forths to determine what is being asked, key suggestions end up being given in Comments.
Double Bonus Wish: move an Answer or Comment to become a new Question, preserving ownership etc..
Reason: it is not uncommon for people to ask respond with questions semi-related to an existing post, but which really should have their own post. (Hmmm, an automatic "Reference:" line would be nice too...)
I wonder if it would be possible to get an editor button that lower-cases the current selection? When people write their question in all-uppercase, it is a pain to re-type their question or to copy-and-paste to an external editor, make the change there, and copy-and-paste back over the original.
When one edits a Question and then returns to the index via the Back button, then if one is sorting by Most Recent Modification, the just-edited Question appears as the first entry of the current page. The system does know the proper time for it and shows it in the correct order if one changes pages, so it is an oddity rather than a deep problem.
I would call this a bug, but not a priority.
Especially when editing long questions, it gets to be a bit wearing that the Save button is (only) below the preview. Could we get an additional Save between the edit box and the preview ?
Downloadable dictionary suitable for IE or Firefox (and potentially for OpenOffice or Word), containing the names of the Mathworks functions, so that when we type in MATLAB commands in to messages, our browser doesn't complain that we are making spelling mistakes.
My iPad is slowly learning MATLAB. Which is actually kinda cool. But also slow and painful - "no, meshgrid, one word, leave it alone, stupid autocorrect..."
Somewhere or other in the documentation, we need a note saying plainly that when a Title is edited, the effect is to create a URL which aliases the old one. For example, I had
http:.../4321-j-random-title
and after I renamed it, that continued to be valid in addition to the new name
http:.../4321-john-q-title
This behavior of keeping the old name accessible can be important because URLs may have been created linking to the old name. This is a good feature; documenting it would save people like me from having to worry about such things.
Technically, the dash is not needed either. So, for example, you can get to this question via:
Interesting... the behavior is not quite what I thought it was. It is not that aliases are created, but rather that as long as the number and dash are there, followed by any non-meta character (e.g., excluding '?'), then the match will be found. In the above example, http:.../4321-* would work just as well.
This is, incidentally, a way to link to a question by number without having to remember or copy/paste the entire title.
It appears that answer numbers and comment numbers are unique. If that is the case, then it would be useful if there was a url that we could use to reference them without having to put in the URL that includes the full title.
It isn't any strain to temporary remember a 4 or 5 digit number while one switches to another topic and writes #answer_NNNN or #comment_NNNN as appropriate, but remembering the full title such as 994-wish-list-for-matlab-answer-sections is not practical. It would thus be nice if there was a constant URL that we could use such as (e.g.)
http://matlabcentral.com/answers/jumpto/ that we could then add answer_NNNN or comment_NNNN to, which when activated would redirect the to the proper place.
Or, better yet, add some markup assistance; e.g., <$answer_NNNN text> or <$comment_NNNN text> would expand the $ to a URL prefix suitable for this purpose.
This would be especially useful to refer to particular FAQs such as the homework ones. Though it is true those can be bookmarked and the bookmarked dragged in to the editing session.
As noted above, there is no scroll-bar for narrow browser windows. The lack of scroll-bar turns out to be worse than expected. I noted above that in Answers, content that cannot word-wrap extends past the end of the box on the screen. Well, in Comments, word-wrap does not wrap on as many symbols and the text clips against the edge of the box. The result can be that a lot of a Comment goes missing without any hint except perhaps loss of continuity. If the user happens to think to select the entire comment and paste it into an editor, the entire content is available so this is a word-wrap and display formatting issue. See for example the rather bad clipping here; the comment that follows that was after I manually inserted spaces around operators and manually converted some long integers to exponent notation.
I would call this a bug.
A quoted sentence was not rendered.
Walter mentioned a "Best of Answers" area. A more flexible alternative would be an "Add to my favorites" button, so we can each keep a list of answers that we find really useful.
Of course, we can always bookmark them in our browser, but it would be good to keep them in the same environment as our personal questions and answers.
I get lost in the morass too. Recently I spent fifteen minutes trying to find a particular answer and gave up.
Though the Best of Answers would be aimed more at the incoming people who are willing to do *some* searching but would get lost in the unstructured morass that exists now.
That's an interesting idea, Andrew.
Do we need a "Best of Answers" area? Cross-referencing and tagging and rating the over 3000 Questions and nearly 5000 Answers is probably a hopeless -- but it might be worthwhile to have a mechanism to nominate a Question as a classic question (perhaps to be generalized a bit) with answer(s) synthesized from the various suggestions.
The question, though, is "Who will bell the cat?"
Seriously, though, I agree. Some really good answers are likely to die in obscurity otherwise.
Pretty soon we're going to need a "Best of 'Wish-list for MATLAB Answers section'"! Ironically, if I vote for this answer, people are going to have trouble finding it.
Hi Oliver, I didn't noticed your message until now and wasn't aware that when accepting our own answer there were no added points, ok it's not cheating, thanks for explaining all those things that I really wasn't aware of!
When comments have the solution, it's the commenters responsibility to post it as an answer - if they don't, tough. I agree strongly with Walter that answering one's own question that no-one else can answer is very helpful, and I disagree with Paulo that it is a rare occurrence. In fact, I do it all the time - I re-ask good questions from the newsgroup, wait for a good answer, and when one isn't forthcoming (often) I post my own. This makes it more likely that people will find the answers in future without having to ask the question again. Since you get no reputation points for accepting your own answer, the only reason why people would do this is because they believe it's the most helpful answer - how can that be cheating?
That's not the case 99% of the time I see someone accepting their own answers, see the example above, the OP accepted his answer but the answer it self is just a thank you.
Of course there's a remote change for someone to find a solution after the question but in that case why not just add the solution to the question with edit or make a comment.
Again it's not a matter of points or reputation it's a matter of accepting the best answers, I don't know if someone already made the suggestions but comments could be turned into answers (accept comment as answer button?!) because often they have the solution.
I don't think that's necessarily wrong. If someone asks a tough question and through their own efforts turns out to be the person who answers it, then it makes sense for them to post their answer and accept it. Whether that gives them points or not is a different matter.
Links to "Published MATLAB Files" would help discussions on code and solutions, which include more than a few lines of code.
I often read the "Published MATLAB Files" in the FEX before downloading, e.g. "Comparison of hgsetget and hgsetgetplus"
Thus I want
- Upload "Published MATLAB Files" to Answer
- Insert links in my text
- Replace the URL by a meaningful string
/ per
PS: Will anybody read this in the old thread?
Thanks, Walter, that was too simple for me ;-) I messed around with "href=". Now, I have to come up with a good question together with a "Published m-file" to try my idea.
Per, what I do is open a separate tab, find the information I am looking for there, and copy the URL. Then I switch back, and at the location I want the link, put in a > after the last word that I want to be in the meaningful string, go to the beginning of the first word I want to be in the meaningful string, put a < before that and then a space, and then paste the URL right after the < .