[DISCONTINUED] MATLAB Answers Wish-list #5 (and bug reports)
Rik
on 3 Jul 2019
Latest activity Reply by Rena Berman
on 17 Jan 2024
This is the 5th installment of the wish-list and bug report thread.
This topic is the follow on to the first Wish-list for MATLAB Answer sections and second MATLAB Answers Wish-list #2 (and bug reports). The third started out as New design of the forum - grey on white and the fourth MATLAB Answers Wish-list #4 (and bug reports) is also growing so large it is slow to load and navigate.
Same idea as the previous ones: one wish (or bug report) per answer, so that people can vote their 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)
268 Comments
Time Descending
Now that moving answers/comments is possible, we kind of need these objects to have accessible timestamps with a resolution better than 24h. When multiple people are posting comments on each other's comments-as-answers, it gets a bit difficult to untangle the thread without knowing the order of events. The sequential identifier numbers are useful to some degree, but comments and answers are not comparable in that manner; furthermore, once an object is moved, it's no longer in-sequence.
A lot of other sites will show rough timestamps, but a mouseover will reveal the full timestamp. That'd be fine.
I wish that the Categories feature be much more useful. Check out hte categorization of this Question:
Scroll over the right and we see: Gaming / Historical Contests. If I was actually trying to find this Question, or others like it, I don't think Gaming would be the first Category I'd check in the "Filter By" pane. I've seen a lot of cases where the categorization seems not useful. Calls into question the utility of the Filter By pane in general for finding interesting content, IMO. As an example, this question is also categorized the same way.
I've alos noticed that many questions are not categorized at all. Any idea why that is? I assume that means such questions can't be found via the "Filter By" pane either?
I have been complaining about the AI categorization in other channels. It is often pretty far off. For a while I tagged the Questions with "miscategorized" to flag them, but there were so many of them that I mostly gave up and stopped looking at the categorization.
I've been spending a lot of time browsing old content and making use of the RHS "See Also" related forum links ... and I think they're less useful than expected. The related forum links seem to be drawn from a small pool of frequently garbage questions, many of which are unanswered or answerable. I see the same ones all the time.
I don't know why the pool is so shallow, but it is, and that's frustrating. I'm half inclined to delete some of the unanswerable ones just to get them out of the pool. While maybe cleanup is in order, why is it so persistent at recommending things with no demonstrated interest?
With the search tool and sidebar in mind, the difficulty in finding relevant content is what has made me more inclined to simply provide direct links to things I think are relevant. Even then, I find myself going back and updating answers to add links to newer threads/answers. It's hardly maintainable.
But that got me thinking about the value of being able to have some curation done by humans. I already keep a growing list of links to reference answers myself, and I know I'm not the only one. Sometimes I wish I could just make it public. I have no idea how it would be implemented, and I doubt anyone ever will, but I think it could be incredibly valuable if we could create something like custom public/private "collections" or "playlists" of content (threads/answers/comments).
The idea really struck me when I saw John D'Errico posting a tutorial on doing exponential curve fits, and I realized that there were a lot of good curve fitting examples that I wish I could still find in this giant sea. I think about Stephen23's tutorial threads, some of the "How to ask" threads, and meta threads like this.
Let's say you loaded up the "Why you should avoid dynamically named variables" thread. You look over in the sidebar and there's a little section that says "Part of the Mega Tutorials collection by Stephen23", and there's a list of links to content that an actual human decided was relevant for a particular reason that's probably described by the name of the collection.
Let's say you add an answer or comment to a collection, maybe the answer can have a little widget at the bottom that links to the collection so that a user can just browse it.
I know, I know. It's hardly a well-defined proposition, but the post is long enough already. It's just an idle thought that might provoke some opinions.
I find myself using them as templates. If I already have something written, I will just use my clipboard and paste it back once the quick reply has overwritten it.
I thought quick replies would be great. I do have a couple quick replies set up, but I think I've found that the only scenario that I'm comfortable with a canned response is when I'm probing junk nonsequitor/copypasted answers before shoveling them. Maybe I'd feel different if I were more active on current general questions.
I prefer to either put the links inline in the explanation or at the end of the answer with some context. Quick replies aren't conveniently inserted in a half-written answer.
Sometime a poster posts approximately the same quesiton more than once, for whatever reason. Is there a way for a moderator to merge such duplicate posts, if it is suggested to the mods, and if a moderator agrees? The advantage of merging the questions is that the various answers and comments woudl be in one place for all to see. A possible disdvantage might be allocaiton of answer points, if there are accepted answers in multiple versions.
Sorry if this was already asked and answered. I search this thread and Matlab Answers for "merge" and did not see it already asked.
And I was just throwing another somewhat related editor function in the suggestion hopper while talking of such things...
@Sam Chak, yes we knew exactly what he meant, because many of us have thought the same thing. There was no misunderstanding of what he meant.
Well, I skimmed a bunch of these threads (you might have noticed) and I didn't see anything about this sort of thing, so I'll just throw this idea out there.
I would feel a lot more comfortable if I were able to give some sort of reason or substantiating information when I delete an answer/comment. This usually seems to cross my mind when I'm faced with what I regard as spam in the form of seemingly random chunks of code/text which have been copied and pasted from other comments/answers or from other threads without any added commentary. I'd like to be able to say "comment is a verbatim duplicate of this one" or "this is a partial chunk of code copied from this unrelated thread over here".
I have no idea if there's any review on these deletions or if the "this is spam" checkbox actually changes anything. I think of what it's like moderating spam through the "manage spam" page. If I were to see any of these posts show up there, I would often have no way to know that they were copied and pasted garbage posted on unrelated threads. Since many are copypasta of legitimate comments and answers, initial appearances would strongly suggest they're not spam.
Then again, if deleted items just disappear into the abyss without review, I'd only be satisfying my desire to prove some modicum of reason or due diligence. Any thoughts?
It would be nice as a feedback to the original poster -- That's a good point. I occasionally see things deleted on threads that I follow, and it just says "[user] deleted an answer on [question]". Having that little bit of commentary added to the notification would be nice to know for anyone following the thread.
Speaking of fat-fingering things, I've come close a few times. Partly because I really need a new mouse, and partly because the pages on the site really like to jump around after I've scrolled. Sometimes you end up clicking on something four inches above/below the cursor. Similarly, the recent inconsistent location of the upvoting button has led to clicking on the flag button when I was trying to get a permalink. Maybe I need new eyeballs too.
I didn't know the spam marking fed the AI, but that makes sense. It kind of makes me wary about marking copy-pasted things as spam. I fear I'd be training it to flag legitimate content.
I only really brought up the "manage spam" page as an example of the hazards of decontextualized manual review. I usually try to visit the content in-context and see if it makes sense. ... but yeah, some of them are extremely obvious (and persistent).
"mark as spam" is (automatically) used as input to the AI that (tries to) detect spam.
"closed" questions can be reviewed (and occasionally someone bothers to re-open one.)
The entries in the spam logs that look like legitimate posting content: for those you often have to examine the user name or user profile to see that the posts are intended as spam, such as if the username turns out to contain the URL for a gambling site.
I noticed a minor issue with hyperlink.
Clicking the link (marked with yellow below) in my Content Feed takes me to the wrong position in Question, close to the comment https://se.mathworks.com/matlabcentral/answers/1450984-what-should-go-in-a-next-generation-matlab-x#comment_1993700 near the end of the Question. However, the hyperlink itself is ok, i.e https://se.mathworks.com/matlabcentral/answers/1450984-what-should-go-in-a-next-generation-matlab-x#comment_2066594
Clicking the link (marked yellow below) in the "header section" of the Question takes me to the correct comment.
I checked this issue and a second time an hour later with refreshed pages.
Windows 10 and Firefox. Both are recently updated.
I always thought it just did that because I have it set to expand all comments.
It's annoying, but I kind of expect websites to jump all over the place uncontrollably anyway.
P.S. Yeah, I'm using Firefox.
I demonstrated this to MATLAB (Rena) and she has some theory on the reason. Unfortunately it was not fixed in the new Answers code they rolled out 2 days ago. It did fix the missing/hidden comment bug though.
Actually it's not a minor bug. It's extremely annoying for me. I've found a workaround though. It's particularly bad for very long threads. You end up not a comment or two away but many screens away from where you should be. The workaround is to go up to the top and click on the last activity there are the top of the screen in the discussion thread. If you do that, you should land on the correct location.
It would be really nice if we could double click a function name in our Answers response, click some toolbar icon, and have it convert that word in our response to the URL for the online MATLAB documentation for that function.
Since smart-sense can put in the links, I wonder why it can't do something to automagically highlight them some how as well -- at least with settings here (that I've made no known changes to any defaults I'm aware of, anyway), they're essentially indistinguishable from ordinary text. I think there may be a color change, but it's not enough I can really even tell that for sure. I (sometimes) will use <> to surround, but it's more time/effort than typically take unless pasting a remote link with the CTRL_K route in which case have to edit the display text anyways, generally.
However, the doc link, while useful, doesn't have links to base functionality -- other than functions, it doesn't know about keywords where those could be the most helpful/pertinent to the poster/question raised. I tried to link to <for> the other day -- it didn't know anything about it.
I just type ‘@doc:’ then type enough of the function name to bring up the appropriate function in the drop-down menu (that magickally appears at that point), and click on the one I want. I generally use bold+underline+monospace for function links so I have to go back and do that, however once I highlight the function name that’s straightforward.
A minor issue
By mistake I accepted an old question of mine. I realized my mistake an unaccepted the question. However, in my "Followed Content" both actions are reported as "accepted an answer"
@Rik, Yes, you hit the nail on the head. I edited the question and now "Followed Content" shows accepted, unaccepted and edited.
If the last two actions on a page are taken shortly after each other, the activity feed sometimes displays one action twice. It often goes back to showing the different actions if there is something new.
I have noticed this especially if the OP accepts an answer and posts a comment saying thank you.
PROJECTS is too closely coupled with Git CM
Some users may want to have a limited view of the files in the Set Path folder list.
MATLAB chooses files based on the ordering of the Set Path. Our developers share a CM repository. Some users create temporary projects to do some analysis. They might need only a couple of files in the main project. So to focus on just the files the user needed, files were deleted from the project. Later, after the analysis and the temporary project was deleted, the user pushed his changes to the shared repository. When users pulled from the shared repository, they were surprised to see that some files disappeared.
When I called MathWorks support, I was told that this is by design. When a file is removed from a project, then the file is staged for deletion in Git CM, so when everyone else pulled from the share, the corresponding files were deleted.
But, they said not to worry, because when the file is removed from the project, the file is not actually lost to the file, because Matlab makes sure to keep a copy even though it is deleted from Git CM. So, when other complain about lost files, then the user who removed the file from the project can be added back into Git followed by a Git Commit and Git Push, and everyone can then restore the file(s) by doing a Git Pull. The user will still be able to work fine with the so-called deleted file(s).
This is not good - very ineficient recovery as well as having confusing history. Project is too tightly coupled with Git with respect to deletions. I recommended strongly that when a file is removed from a project, do not by default, mark it for deletion in CM. Or, at least have a toggle where the user can set the default to not to delete it from CM.
In 2020, after a short tryout, we had to stop using projects.
Background: I recently put some effort into answering How to store and reuse coefficients in a for loop. My answer together with a comment includes a discussion on "column-major order" and "performance". I searched in vain for references in the documentation and in Answers regarding the significance of taking "column-major" into account. OP has read my answer an comment. So far so good. Now I would like a few more users to read my responses. To make my responses to OP's comments more visible, I sometimes add them at the end of my answer. And I add tags, in this case I added "column-major"..
Proposals:
- support multiple code blocks/wells in one answer. In the current case my answer contained one code block and my comment another. Adding the comment to the answer causes "Run in 2021a" to throw an error for the second code block.
- support searches (Cntr-F) in the current thread to include hidden comments. Currently, "Search Answers" for "tag:"column-major" singleton" returns the question ("singleton" is in a hidden comment), but Ctrl-F doesn't find "singleton". (I use Firefox.)
As a temporary work-around you can do what I do: make use of the fact that the output is persistent. I sometimes reorder everything to make it run with the correct output in the correct places, and then edit the code to a more sensible structure. One example was a user posting a question about fminsearch where their cost function didn't return a scalar. Easy enough to show, but I didn't need the entire function to do that, so I just ran it with the function and then edited away the function, leaving only the call.
So if there are plans to remove this persistent behavior, please don't.
Sign in to participate
