Posted 28th May 2020 at 13:59Just a small update:
* I've now restored kakuro puzzles to the rotation
* For moderation purposes, I now have the ability to lock comments so they can be no be longer edited
* Similarly I can now lock entire topic or puzzle comment pages, should I need to

Gareth, puzzlemix

Posted 28th May 2020 at 14:04
Last edited by gareth 28th May 2020 at 14:06In terms of future plans, I'm also planning to update the puzzle completion window to fix the behaviour on mobile and tablet devices, since iOS now completely blocks the pop-up completion window without any indication at all (unless you go into Settings->Safari and enable pop-ups - which also doesn't open it, but at least gives the user the choice of viewing it). This will then affect all users, since I'll change it universally, to open within the player window rather than as a separate window.

I also have wondered recently if the huge preponderance of 'easy' ratings is because to many users it will appear the quickest way to shut the completion window is simply to rate it 'easy' first, then click 'close', so I'm going to remove the need to choose a rating before clicking 'close'. If this works, I'll then retrospectively scale up all previous average user difficulty ratings to allow for this bias.

Finally, I'd like to restore the solving aids for touch-only devices, which currently are hidden in the UI for no particularly good reason.

Posted 28th May 2020 at 14:07If there are other update requests you have, in terms of basic functionality for the site, please feel free to post them here - even if you've asked before.

Posted 28th May 2020 at 22:33Great idea to collect suggestions in one thread.

comment:
I almost never rate (I think I wrote about the reasons recently in the comments) and do not so by just clicking the X in the top right corner in my Windows based layout's pop up window. Still works fine just a minute ago...
suggestion:
Put down a button for each rating, close the window if any button is clicked.
If it is included in the player window, please keep allowing not to rate.

suggestion:
In this recent puzzle
https://www.puzzlemix.com/comments.php?id=87245
I had trouble with the two yellows - maybe the two pinks might be difficult to keep apart, too (not in this one because the only possible mix-up is already a "C"
I guess the color order is based on the current number of the area according to first appearence, so in a Wraparound Sudoku it might be not possible to avoid this problem... could be a reason to call it more difficult.
(and I think there was a recent request by someone else about the colour of th X-cells compared to yellow cells.)

suggestion:
When a solution is correct, the puzzle is locked after clicking the "check puzzle" button. I'd like to be able to unlock it to backtrace (Ctrl-Z)

Thank you!

Posted 29th May 2020 at 01:27
Last edited by gareth 29th May 2020 at 04:36Well, I’ve now rewritten the code for puzzle completion/giving up.

Works for me - but please let me know of problems on other devices!

Included with this I changed three other things:
* You no longer need to rate a puzzle to have a clear 'close' link
* You can jump straight to the stats/comment page for a puzzle on completion
* You can't comment on completion - you have to go to the comments page

There is also a minor change no one will probably notice, which is that the full stats don't update on an open underlying puzzle list page - previously all the stats updated as you finished; now just the play status and play count updates. This really doesn't matter, however.

Posted 29th May 2020 at 04:33Fixed a few things:
* For a couple of hours the puzzle play counts were resetting to 0 whenever a puzzle was opened - now fixed
* On iOS/iPadOS, you couldn't use the rating controls on the non-number-puzzle puzzles - now fixed

On Firefox on PC (and no doubt other platforms) I can't get the difficulty rating selector to work in the player window. I'm not sure why yet - an event must be being trapped that I need to release for it to function, but I can't work out what. So if you want to choose a difficulty level and it doesn't work, and you're using Firefox, please try another browser temporarily. :)

Posted 29th May 2020 at 12:23
Last edited by kocos 29th May 2020 at 12:43Thanks Gareth, great changes.

Maybe a number of ratings could be informative in the average rating column. Like this: Mild (23)
Have it added it now, or was it always there in the statistics page?

Posted 29th May 2020 at 12:58As somebody else previously suggested, could you put a small border at the top of slither link puzzles so that it is easier to place lines when playing it on a phone or tablet without accidentally hitting the browser buttons. Sometimes it won't even let me place lines up at the top without zooming in. Many thanks!

Posted 29th May 2020 at 14:56I was reviewing the player code last night and interestingly it deliberately removes the border on some mobile devices. I'll just change that so it never does that now - I imagine this was to cope with the way an older browser version worked, probably on iOS.

I need to update to make rating in Firefox work too - no idea what the issue is, having spent a long time trying to puzzle it out, but I'll just redesign it to have three pictures to click on instead and hopefully that will then work.

kocos, that's been there for 13 years. :)

Posted 29th May 2020 at 16:18
Last edited by JoergWausW 29th May 2020 at 16:18"previously all the stats updated as you finished"

i remember it this way: When I finished a puzzle that I had started from the comments.php page, my personal stats in the table were updated, but the section "your puzzle statistics" never was updated. I had to refresh the page for that. I don't know for the rankings.php page.

"that's been there for 13 years. :)"
Maybe the idea was to also give the numbers for the ratings that differ from the calculated average. If it says "easy (36)", it may be interesting how many people scored it harder. Or do I miss this information? (of course you could count yourself in the rankings...) What is calculated here? An arithmetic mean or the median?

Posted 29th May 2020 at 16:25
Last edited by JoergWausW 29th May 2020 at 16:25About the Firefox issue.
I just took a quick glance at the code of your selection form there and thought about the "onChange". I thought that it looks funny and googled this:
https://stackoverflow.com/questions/4784262/is-onchange-event-case-sensitive

Maybe Firefox is choking on that C ?

Posted 29th May 2020 at 17:46
Last edited by gareth 29th May 2020 at 17:50Hi Joerg,
The averages are all means.

Thanks for looking at the code. I just tried changing that (on my local copy) but it had no effect unfortunately. That is also cut-and-pasted from the previous completion window, so it's been like that since the site launched - which is not to say Firefox might behave differently in different contexts.

I presume if I swap for direct links it will be fine, but I'm curious as to why it isn't working. The page responds to various events during solving, but I believe I have disabled this on completion - which I needed to do for the non-number puzzles for the dialog to work on iOS for a start. So quite why the remaining problem only affects Firefox I don't know. Maybe there is some ancient Firefox-specific workaround somewhere in the code that is causing the trouble. There is a lot of unnecessary code for old browsers, since the player still works back to IE 5.0 (but without saving and, since I modified the code yesterday, storing completion status), with full support for IE 5.5 onwards. It even supported the weird IE 5 on Apple computers that Microsoft messed about with for a while, before abandoning. Might have broken that in the intervening years, but I doubt it.

Actually, since the site switched to https that has probably done more to break really old browsers than anything else, so the above may be true re the HTML but not once you allow for the transport.

Oh, and that's right - the statistics on the comments page never updated, just the horizontal bar at the top, i.e. the same as on the main puzzle menu. But now if you click the link it will navigate the window, or its parent if still open, to that page and that in itself causes a full refresh. During play it does still update the play icon correctly, though.

Posted 29th May 2020 at 18:02Just googled some more, found this:

https://support.mozilla.org/en-US/questions/1150392

I didn't check any suggestions of those.

I just tried to see if arrow keys work. Nope.
Then I noticed this:
Hitting the tab-key, there is always a dotted box showing what input tag I am at. This doesn't move away from the "close window" button. Maybe you force firefox to not allow anything else but this?

Posted 29th May 2020 at 18:07
Last edited by gareth 29th May 2020 at 18:13It's now fixed, thanks to an email from a user - thank you!

There was one event I missed when I was deactivating them, which I've now covered too and it's now working in Firefox. Yay!

I have also changed the iPhone and Android player margins (in a not very scientific way), in the hope of fixing an issue with too-narrow margins above Slitherlinks on Android. Anyone playing on Android please let me know if this has helped/made it worse. I don't have a modern Android device to test on.

Posted 29th May 2020 at 18:09I just did the recent Nurikabe and wondered that it suddenly worked :-)

Posted 30th May 2020 at 03:34I've also now changed the way the scale for the 'completion' (or 'gave up') dialog that appears above the puzzle works - it now should be entirely consistent in height, rather than varying for extra-wide puzzles.

Any remaining problems, please let me know. :)

Posted 30th May 2020 at 10:21Just tried a slither link puzzle on an android phone and I can now add lines to the top row without having to zoom in. Much better, many thanks!

Posted 30th May 2020 at 18:11
Last edited by gareth 30th May 2020 at 18:18Thanks rh5528 - good to know! I've had to adjust the code again, but this shouldn't have affected Slitherlink much (I slightly shrank its maximum size but left the top margin the same). On Android hopefully the other puzzle types no longer overlap the edge of the side panel now, however.

In other news, I've also fixed an issue with the larger demo puzzles on the front page not displaying if you weren't logged in (oops), adjusted the completion dialog when not logged in (much briefer now), and cleaned up the link display in the completion panel. I also fixed the 'add comment' box for forum posts if you aren't logged in, which let you type in a message even though there was no submit button. :>

Any remaining issues, please let me know!

Posted 30th May 2020 at 20:17
Last edited by JoergWausW 30th May 2020 at 21:34I thought I already commented this issue - but there is no function to review all my recent comments, is there?

issue:
When I finish a puzzle nothing is updated on the comments.php site. When I start the puzzle the icon changes though.
Update: It only updates if I click "close window and show stats".
Added: The check solution dialog doesn't say anymore what my previous time was.

little request:
is it possible to get the current played time shown when the "check solution" result is wrong?

Posted 31st May 2020 at 02:55
Last edited by gareth 31st May 2020 at 03:10The player changes are by design to avoid complexities with varying heights - although I’ve just found it’s too tall on iPhone even so.

The time should update in the comments and menu pages on completion - I seem to have broken that, so I’ll look at it soon.

I feel seeing your time while solving adds an unpleasant pressure, so have no plans to add that feature. It could be an option, but I’ve tried options before and no one - literally no one - uses them. But if it becomes a popular request I’ll add it! :)

And there’s no search option for comments. Not sure it would be useful? Maybe seeing your own might be useful though.

Posted 31st May 2020 at 09:27about comments: I just wanted to look for my own comment, because I thought I posted it, but couldn't find it.

I don't want to see my time while solving. You are right about the pressure argument. I'd like to see my time after I hit the "check puzzle" button - even if my solution is wrong. That way I can tell afterwards how long it took me to find my mistake.

Posted 31st May 2020 at 11:34I tried today's slitherlink on my phone, and it worked much better than before. Previously, I had the most problems with the left edge (not the top edge), even when zoomed in, but today it was ok. The grid was too large to play, so I stil had to zoom in, which made the puzzle more difficult. The time showed it. But then I was much slower on laptop, too, so I could have been the lack of coffee.

Anyway, I got a small request: when playing on the laptop, I prefer using the keyboard (slitherlink, hitori, and nurikabe puzzles notwithstanding). I would like to finish the puzzle by pressing some dedicated key instead of having to use the mouse to click on the "check solution" button. Would save me a couple of seconds. :-)

And a slightly bigger request: I often go over the puzzles a few more times to see where I could have made better choices in order to solve the puzzles faster. For sudoku-style puzzles, highlighting the pencilmarks for a selected number is sometimes useful. Either for finding singletons in a large 16x16 grid, or detecting more complex patterns like "swordfish".

Another solving tool: highlighting incomplete pencilmarks on request (i.e., pencilmark does not contain the correct value). Could be useful for complex sudoku (killer, wrap-around, etc), especially for large grids.

Posted 31st May 2020 at 13:24
Last edited by gareth 31st May 2020 at 13:26Hi JoergWausW,

Interesting idea - I'll think about that!

Time is quite a tricky thing to deal with in general, since you need to allow for the odd person who wants to "cheat" their time, and I don't really want users to have to click "start" at the beginning of the puzzle and have them wait for the server to authorize their start time - thus the current behaviour of starting the timer when the puzzle first fetches from the server. This fails occasionally when people 'double click' open the puzzle and the server thinks they have started twice, although I could 'debounce' this now I think about it since two clicks close together can be merged into one. But the vast majority of the time it works fine, and it is robust. But there's no reason why the client can't read the server time and display it at appropriate points, as you suggest.

I also once thought of recording solves - which the player does anyway during solving since you can undo back to your first move - and making them available to watch back. Those could be timestamped if so too.

Although in truth I think any major development should be saved for a new client, since the player is 15 years old and could do with being rewritten for modern web technologies. My general plan is to write a generic player which just deals with an abstract canvas, and can which then be fairly easily be ported to iOS and Android native apps. However when I do that I will start with iOS and port to the web later, since the current client obviously works just fine as it is - but is unnecessarily complex to develop new features for.

Posted 31st May 2020 at 13:34
Last edited by gareth 31st May 2020 at 13:36Hi winterer,

Thanks for your suggestions - all fantastic ideas!

1) The keyboard thing is easy - my only concern is whether people might press it by accident and get confused. What key would you suggest? We have Z and X for undo/redo (no CTRL needed), and I think S should be for save, plus A to F are for playing 1-16 sudoku. So maybe Q, but that's a bit close to the number keys. I don't want to use return/enter for it.

2) Highlighting all pencilmarks for a given number wouldn't be too hard to implement, but how do I add it to the UI without causing confusion? I could hide it away with a SHIFT+number keypress or something, but then it becomes impossible to ever make available on mobile - although currently on mobile I have hidden the options anyway.

3) Highlighting pencilmarks missing values - interesting idea, but this only works for players who have specifically filled in all pencilmarks, which many don't - so it would need to be an option specifically called 'Check pencilmarks' and again how do I add that without causing confusion? I've found people don't even read what's on the player already, and never venture to try options - even skilled puzzle players who play other online games have said to me things like "I wish I could place pencilmarks" to which I just think how much more obvious could it be? :) So the more I add less-used features the harder I make it to find the more-used features, if you see what I mean. Fundamentally, of course, this would be relatively easy to add too.

Perhaps the player should have a 'solving aids' panel that appears when requested, and all the current options should go into there too. That could then be a bit more complex, since it would only appear when wanted, rather than by default.

Posted 31st May 2020 at 15:57- about time, I sometimes have this situation: I click the puzzle and then the "loading puzzle" message is visible for 15 secondes before the puzzle actually shows. That's due to a slow internet connection. So "first fetching from the server" is not the same as delivered. To prevent a time disadvantage, the player tool would have to send home the message when the puzzle was completely recieved... but then, again, the "cheater" might be able to delay that message....

- for finishing the puzzle something like ctrl+shift+enter would be easy enough excluding accidental hits (I'd like to have that short cut option, too). Tablet users don't need it anyway, because they can just touch the button. Or the "end" key might be off enough and a good association.

- I'd also like to have the keys for the letters DEFG next to each other for convenience. The E is always stupid to get to. Maybe it is possible to make an E also appear/disappear when you press R (it's kind of between D and F), of course the E key should still work.

- another feature, that people might like to use: save a current puzzle state constantly with a specific file name and an option to publish it to the comments. Sometimes people like to ask for help when they are stuck, others like to help. I store those pictures on my private server, but not everyone has that option.

- the idea of a solving aids panel sounds great.

- about a missing pincilmarks option: if the user doesn't set any pencilmarks and uses the option - will it show the complete solution automatically? Because the program can only know valid pencilmarks from the solution. If it is supposed to calculate valid numbers by current puzzle state that might be tricky to do, because what is still valid always depends on the knowledge of the player... (I'm afraid this tool won't work)

- another thing about pencilmarks that confuses me: If it is a 7x7 Futoshiki or a 8x8 Sudoku I'm still expecting the 6 in the second row. Of course it looks more symmetrical the current way...

I'm sorry for my long list of suggestions, Gareth. Your puzzle player is already the best I've seen so far. But you asked for it....

Posted 31st May 2020 at 16:00Hi Gareth,

1) this should be a shortcut to the "check solution" button. On mobile, there's no large speed penalty, as clicking the button with the finger does not involve a switch of input device, so to say. On a keyboard, you could add a short-cut like CTRL-Enter. I would make it a spcial key, using CTRL or ALT or SHIFT, as a safety measure, to decrease the likelihood of it being pressed accidentally.

2) here, mobile UI does become an issue, which I had overlooked. However, as this would count towards using solving aids, I probably wouldn't use it during a timed first solution attempt. So activating this via some button, i.e., clicking "highlight pencilmarks", then the desired number key, would be fine.

Side remark: on a phone display, the buttons for "check", "save" etc are quite close together. Sometimes too close, especially to the lower row of numbers... :-)

3) Yes, same as for 2), could be triggered by the "check pencilmarks" button you described. I know this feature form another site, where it's sometimes helpful for extremely tough killer sudokus (9x9s that take upward of 30 minutes to solve), to find the last point where my pencilmarks were still in order when extricating myself from a dead-end in the search tree. Note that empty fields (no pencilmarks at all) would not be marked as incorrect.

However, featuritis, the curse of all software that grows old. My own code suffers from it. So if I could have only one wish, it'd be 1), the "check solution" shortcut. The other two I would use rarely, only on particularly frustratingly complex sudokus, like some of the 12x12 wraparounds. Or the recent 16x16, where I lost 10 minutes because I didn't see a singleton 2. So the development effort might not be worth it after all, for the few times a single user might use it. Or save it for the "solving aids" panel in the new client, if and when you get to coding it. :-)

But even the check short-cut might not be worth it, as it saves a couple of seconds at best. At the start of the puzzle, another click is required to bring the focus into the grid, while the time is already running. But this is only a problem for the puzzles that could be solved faster with the keyboard, so a distinction by puzzle type is needed... This gets more complex the longer I'm thinking about it. Just to shave off a few seconds... So, in the end, I think I'm fine, thanks! :-)

Posted 31st May 2020 at 16:09Hi JoergWausW,

"about a missing pencilmarks option: if the user doesn't set any pencilmarks and uses the option - will it show the complete solution automatically?"

No, the tool that I had in mind only checks if the pencilmarks contain the correct value. If they do, nothing happens, if they don't, the pencilmarks in these fields are highlighted (red, in the case of the website that I had in mind).

The correct value is not added to the pencilmarks. (this is already covered by the "auto-add valid" button!)

Empty fields (no pencilmarks) are not highlighted. (well, maybe they are, but as they don't contain any text, there's nothing to be coloured red).

Posted 31st May 2020 at 17:53Quite like/don't mind the new way of rating/closing BUT when you check ratings and then go back to the puzzle list it resets the filter "unplayed" which is slightly annoying. I am using Google Chrome if that helps.
So excited to see new Hanjie sets.

Posted 31st May 2020 at 20:04I don't know if anyone's already commented on it, but I'm just saying it just in case. If a puzzle is done a second time, ask for "choose your rating". This didn't happen before. The same is true in "permanent sets of puzzles". That is, a player can change the rating of the same puzzle.

Posted 31st May 2020 at 21:25
Last edited by gareth 31st May 2020 at 22:36Hi everyone,

A few short replies to the easy bits. :)

* Thanks carlotes - as it happens, you’ve always been able to change difficulty ratings on replay. Previously it would auto-select your previous rating, but you could change it. So perhaps it did appear you couldn't change it before - but in fact you could.

* Thanks also catkinpuzzlemix - if you open the comment window from the menu, then open the puzzle from that, it never updated the (now closed) menu, so that hasn’t changed. It can’t if the window isn’t open any more - but is that what you mean? If you go back the browser may then retrieve a stale cache of it (i.e. an unchanged page), but I can’t change that unless I make the page poll every few seconds, which I don’t think is a good idea. There are other ways of structuring things, but fundamentally unless a page proactively polls for updates it’s not possible to cover all windows beyond the parent or its older direct ancestors. There may be a way to mark the history as stale from the comments page, but I’d have to look into that and see how well supported it is across browsers - you can definitely do some history manipulation in modern browsers, but I’ve never tried that.

Posted 31st May 2020 at 22:31In terms of the menu window icon not updating, when you go back to it having opened the puzzle from the comment/stats page, I can't actually reproduce that problem catkinpuzzlemix. It does auto-refresh. If it didn't, I think I could use History.replaceState() to force the page to refresh, but that doesn't seem to be an issue?

Can you let me know exactly what you are doing, in terms of where you start from and how you open each window? I'm using Chrome too. Where do you start from, what do you click on and with which button - e.g. are you opening the comments in a separate tab to the menu? What are you closing and how are you going back to the main page?

Posted 31st May 2020 at 22:33
Last edited by gareth 31st May 2020 at 22:33Winterer, that's what I thought you meant - although the idea not to highlight squares without pencilmarks at all is a good refinement to what I had imagined.

I'll think about what changes to make to the solving tools, but next time I update the player (and I want to do so asap, to restore updating your time in the parent window - and fix your time being partially obscured on iPhone) I will add a keyboard shortcut for completing a puzzle.

Posted 1st Jun 2020 at 19:06Finished a puzzle clicked on easy then "Close window and see ranking(statistic?)" at which point the Puzzle staes filter changed from Unplayed to All puzzle states.

Posted 2nd Jun 2020 at 10:25
Last edited by JoergWausW 2nd Jun 2020 at 10:25minor glitch:

menu.php in the box "Permanent Sets of Puzzles"
There are some 4-digit-numbers shown where there was nothing before, I think it is some blank text if you didn't purchase a text and it is not a new one...

Posted 2nd Jun 2020 at 12:19JoergWausW, thank you so much for pointing that out! A bit of temporary debug code I forgot to remove, as I was initially wondering why the "New!" didn't appear on the Hanjie sets I added at the weekend! Fixed.

Posted 2nd Jun 2020 at 12:24
Last edited by gareth 3rd Jun 2020 at 01:05So here is my summary of what needs fixing:

* Make time visible in player on iPhone/iPad on completion - currently part-obscured
* Preserve menu sorts and filters when clicking 'view stats' link from player (currently view settings are then forgotten)
* Update time on completion in menu/comments window - wherever player was opened from

And things to do in the short-term:
* Add a shortcut to complete the puzzle - will be Q after all I think (and to save, if there isn't already one - I forgot if there is - but should be S)

Ideas for other improvements:
* Add extra solving tools discussed above
* Move solving tools into a new panel, and make accessible on mobile/tablet

Posted 3rd Jun 2020 at 01:43
Last edited by gareth 3rd Jun 2020 at 01:44Okay, I believe I have fixed all three issues above:

1) I've made sure text is visible on iPhone and iPad (at least on the devices I have - please report any remaining problems)

2) The puzzle and puzzle state choices are now preserved across opening the player and clicking within it to view stats (other choices already were, since they are in cookies - but those weren't since I had discovered previously that they caused great confusion by maintaining their state once selected!)

3) The time now updates in the parent opener window on puzzle completion, whether the menu or the comments page.

In the process I also restored the 'previous best' to the results, since I altered the layout slightly to move it up top.

I've also added 'Q' (Quit) for 'Check puzzle', and 'S' for 'Save puzzle' as keyboard shortcuts. This is shown in the button labels too, if you're not on mobile/tablet.

Three other shortcuts I'd forgotten existed, but I rediscovered, are that you can use / * and - on the numeric keyboard (above 789) to be A B C when solving a 12x12 (or larger) puzzle.

So this now completes my changes to the player, at least for now, I hope.

Posted 3rd Jun 2020 at 08:40Gareth,

Thanks for the updates. Though Q is a dangerous choice, as it's really "a bit close to the number keys", as you first wrote.

As for solving aids, here are some that I sometimes would like to have available, in (assumed) increasing order of difficulty of implementing them. Some would be quite easy to implement if a good CP solver (constraint programming) was available in JavaScript (which there isn't, to my knowledge). For others I don't even know how to start to model them yet.

So I don't actually expect to have any of those available in a JavaScript-based player anytime soon, or ever :-) :

- Calcudoku: let "remove invalid" remove all pencilmarks that cannot be part of a valid instantiation of a given region. E.g. a two-square region marked "24x" would have only pencilmarks [3,4,6,8]. If, e.g., the 4 gets removed in one square due to its being placed in another square in the same row or column, the 6 gets removed from the other. (i.e., achieve generalized arc-consistency regarding all constraints on the regions)

- Mystery variants: the same as above. E.g., a two-square "15?" would retain pencilmarks [3,5,6,7,8,9]. Updates as above.

- Kakuro: same as above regarding upates, i.e., generalized arc-consistency when using "auto-remove". E.g., in a run with clue "16", when the pencilmarks of one field is reduced [9], the pencilmarks of the other gets reduced to [7].

- Hanji: shade cells light/dark if that follows necessarily from current user input and the possible distributions of the black cells according to the given clues. E.g., in a 5-square row with clue "3" and second cell marked black by user, shade third cell dark and fifth cell light. Analogous to "auto-remove invalid" in Sudoku.

- Slitherlink: mark cells that close a loop. Maybe colour closed loops.

- Hitori: mark cells that would cut off a region. Maybe colour cut-off regions (light grey)

- Nurikabe: reduce the count in the clue cell by the number of white spaces connected with it.
Mark cells that would cut off a (black) region. Maybe colour cut-off regions (dark grey)

- Slitherlink/Hitori/Nurikabe: generally shade cells automatically dark or light, if this follows necessarily from current user input (as, e.g., described in the example in the Slitherlink rules). Analogous to the "auto-remove invalid" in Sudoku. Obviously, this depends on the complexity of the global constraints in the model and the difficulty of achieving generalized arc-consistency (in CP speak). Ideally, only remove choices that are directly invalid and don't let this propagate until a fixpoint is reached (this is unlike CP then), as otherwise this would (almost) become an auto-solver

However, what I would really like to have is a strong CP model for Nurikabe, if this is indeed possible - this is the one puzzle type where I am still quite lost at sea. If I can model it, I can solve it efficiently. But the few constraints that I know of are rather weak, so there is still too much guessing...

Posted 3rd Jun 2020 at 17:04Hi winterer,

I don't know whether it is allowed to post links to other webpages here (if not, gareth, please feel free to delete this whole paragraph): https://www.sudokuwiki.org/sudoku.htm

I found this Sudoku solver page that uses a lot of techniques in order of difficulty. So basically, what you want, is a panel with solving aids, that works like that...?? As you said, not a job quickly done.

So far I didn't take the time to study all those methods, and this is probably, why my times are slow...

Once upon a time, I did well with 7x7-Skyscrapers, then I talked about my method here in the comments, and now I'm behind... at least I seem to have a slight advantage with big wraparound sudokus and toroidals. So please, more of those!

So I guess, in the end, part of the fun is to figure out the best ways yourself...

@gareth: what about those sum skyscrapers? I saw that you tested it...

Posted 4th Jun 2020 at 00:02It's fine to link to sudoku-solving sites, and thanks JoergWausW.

winterer, you can already do pretty much exactly that for kakuro and killer sudoku 9x9 - if you enable the 'show current clue fits' option it will cross out options that no longer fit based on your pencilmarks. There's no reason the player can't do this for all types of number puzzle - I just never got round to it. It creates a static lookup table before the page loads, but it's a simple recursion to calculate all possibilities. With e.g. calcudoku, however, there might be UI issues for large regions with many ways of reaching the total.

For slitherlink, if you forget about automating applying the result of a deduction, you can also do most things already - the only one thing covered is whether the loop is connected. Everything else will turn green or red with 'show valid/invalid placements'. Similarly in Hitori, it's only the disconnected areas that aren't highlighted by the tools.

On the shading puzzles, it would be helpful to be able to show WHY a solution was invalid in ALL cases, so the most important features would be:
* disconnected areas in hitori / nurikabe
* incorrect numbers in nurikabe and hanjie
* multiple loops/lines in slitherlink

If everything was covered, a logical inconsistency could then be automatically highlighted when you 'check' the puzzle, which of course is distinct from 'show wrong'.

Tools to highlight errors are the way to go, since it becomes impossibly confusing if you start auto-adding solved segments to puzzles. Removing pencilmarks in sudoku is a bit different, since you don't tend to 'sketch in numbers' in sudoku - whereas in slitherlink/hitori/nurikabe you might try some things out as you experiment.

Posted 4th Jun 2020 at 00:05On solving Nurikabe, once you've made 'obvious' moves then make sure you always cycle through the constraints in your head:
* Is there a 2x2 area only one (or two) regions can reach?
* Are there cells this number MUST visit
* How am I going to keep all shaded squares connected?

Then try things like:
* Are the numbers so big in total that they all more or less must fill all possible space around them?
* How do the corner squares connect?
* Must my shaded area flow around the outside of certain numbers?

Posted 4th Jun 2020 at 01:05
Last edited by gareth 4th Jun 2020 at 01:09New update, to deal with being offline when you finish a puzzle:

* If the save/time-lookup on completion fails, you now get a chance to retry (as many times as necessary)
* Also, if the completion server connection doesn't respond quickly, after 4 seconds you get an acknowledgement that it's waiting, and after another 4 seconds you get the option to retry manually (it will still update successfully if it happens to succeed after 8 seconds)

I've also renamed this topic to better indicate what's in it!

Gareth

Posted 4th Jun 2020 at 01:13
Last edited by gareth 4th Jun 2020 at 01:21So, just to summarise potential future improvements again, and to bring this up to date:

* Add extra solving tools discussed above
* Move solving tools into a new panel, and make accessible on mobile/tablet
* Fix colour assignment in jigsaw/wraparound puzzles so like-colours don't touch
* Extend 'clue fits' behaviour for calcudoku and killer plus puzzles, as well as killer sudoku that aren't 9x9

Other issues:
* Make using 'show repeated numbers' a 'major aid' in Hitori (or disable removing underline once shaded)
* Fix the comments page to not show "-1" sometimes for one stat after solving Hitori

Other ideas I've had are:
* User profiles - possibly a brief option to add an 'about yourself', but mostly to show stats on your site usage (e.g. puzzles played, puzzle plays over time, etc)
* Badges and achievements, which might also show to some extent on comments (e.g. "[] More than 1,000 puzzles solved")
* Further privacy controls for the above two options
* Delete user account button (to save me having to action these when I am emailed)
* Option to view all your comments, all replies to your comments, and all new comments since you last visited the discussion forum page

One other thing I'd love to fix, but don't know how to:
* Work out how to get rid of the giant toolbar on an iPhone, when solving in landscape mode. In previous OS versions I could do this, but it doesn't seem possible without scrolling the window now, and it comes back very easily when you go near the top of the screen. This isn't my greatest priority, so not worth a disproportionate amount of effort.

Posted 5th Jun 2020 at 11:07The puzzle state issue seems to be fixed, thank you.
Have also often wondered whether there was a less tedious way to shade large areas/rows etc in Hanjie. E.g I've tried holding the shift key as one might do when highlighting multiple emails to clear an inbox but that doesn't work. Any ideas? I'm sure there must be something I'm missing otherwise I have no idea how people shade all the boxes in a large Hanjie in just a few minutes - and that's without solving it!

Posted 5th Jun 2020 at 16:15
Last edited by gareth 5th Jun 2020 at 16:18Hi catkinpuzzlemix,

I'm glad the puzzle state issue is now fixed for you!

For Hanjie, just hold down a button and drag the mouse - or drag a finger if on a tablet/mobile. If you're clicking one by one on each individual square then I can imagine that would take a while!

Dragging works for all the shading puzzles, and it's massively quicker.

PS This is explained at the bottom-right of every single Hanjie puzzle on the site, in the text at the bottom-right of the player. This also explains what SHIFT does - it toggles a state, although I don't usually find that useful myself. You can also click the links (or click 'Instructions' at the top of any non-player page) to view full instructions on how to use each player. There's also a very large piece of bright-yellow-highlighted text at the top of the player every time it opens pointing you to that text, since I found long ago that people didn't tend to read it.

Posted 8th Jun 2020 at 18:17No new Hitoris for a week. Are we out again? :-(
Or is it just a random improbability? Did Kakuros take over?

Still waiting for a set of nice and big ones of Hitoris... (are these also hand made like Nurikabe?)

(Maybe some internal function to alert the admin for new supply is needed?)

Posted 8th Jun 2020 at 21:07There are still a considerable number of prepared hitori, so it's just coincidence. And I don't need a system while you're still playing! :) But yes, I suppose adding the kakuro in again did make hitori less likely to appear.

Posted 10th Jun 2020 at 13:01Hi Gareth,
Thanks for all the updates. I like the 'Q' to exit a game rather than needing to find the mouse and get the cursor to the right place.

One suggestion that I have. When moving the cursor with arrow keys, I don't understand why when you move off the right hand end of the grid the cursor reappears one line lower on the left hand side of the grid. Why does it not stay on the same horizontal row? Other sites have it that it stays on the same row so it is confusing at times.

Thanks again

Posted 10th Jun 2020 at 14:42
Last edited by gareth 10th Jun 2020 at 14:42Hi TranslatorBJ,
Glad you find 'Q' useful, and interesting comment. I designed it like that in 2005 and haven't thought about it since! But, thinking about it now, I feel I do often go off the right and then immediately press up, so perhaps that should change. You can still press 'tab' to move to the next field if you want anyway. In fact, I suspect I simply aped the behaviour of a very primitive player I had in late 2004 that used a form where you typed in fields and had to tab to the next box.

Okay, I agree. I will change it next time I update the player. I will keep 'tab' with the existing behaviour (and note you can also shift+tab to do the reverse), and modify the behaviour of the right and left arrow keys at the end/start of a row respectively.

Posted 10th Jun 2020 at 16:54Actually, I see advantages of that behaviour and can imagine asking for the same vertically: to move one column to the right when going off at the bottom of the board.

I often fill in pencilmarks for all nine 3x3 boxes first (without the invalid). After that I start in the top right corner and eliminate the invalid numbers that are in that current row. I just have to use the right key and can walk through the whole puzzle, finishig in the top right corner. Then I continue walking down vertically and I have to do it differently: I always have to move the cursor to the next column by hand .

One idea is to unify the behaviour, The current state would be a compromise...

Tough topic.

Posted 11th Jun 2020 at 13:59New Hitoris - again, not today either :-(. It's been 10 days now. Before there was one every other day (I don't really believe anymore in random reasons)

Gareth, you said, there is a check that no puzzle type appears on two consecutive days. Maybe it is possible to introduce another check that a puzzle type has to be published after one week without?

Posted 11th Jun 2020 at 16:05There's a Hitori coming up tomorrow. :)

Posted 11th Jun 2020 at 16:44How do I turn of the shortcut keys Q and S?

I accidentally hit Q frequently and am told the puzzle isn't finished. Which I know. At that point any solution I post won't be shown as solving it in one go with no help.

I'm clumsy with keys perhaps and it's right by the 1 and 2. And I know from other comments that some people appreciate it because it's quicker than spending two seconds using a mouse. But I'd really like to remove it. Really, really like that.

I need to not have that shortcut. I don't need any shortcut. I have a laptop and so have a mouse. If there must be a keyboard shortcut why not stick it away from the buttons you'd be pressing for the puzzle? Or add a shift or alt button so it can't easily be pressed accidentally?

Thank you.

Posted 11th Jun 2020 at 17:48Hi ClareJuliette,

Sorry it's causing you problems. I will address this next time I update the player. In the meantime, can I suggest you use the number keypad to enter numbers, so you are then a very long way away from the 'Q' key. I hope that helps for now!

Gareth

Posted 11th Jun 2020 at 23:29I guess there is no number keypad - and that's the problem... For me the Q is too far away from my number keypad...

Back to topic: Why not use the End-Key? Or Escape? Wait - even further away than Q... Nevermind.

When I bought my laptop, I didn't care about size, so I bought one with a number keypad.
Recently my display's hinges broke because of it (too heavy display for plastic hinges)....

Posted 9th Jul 2020 at 17:13Today, I clicked the puzzle link on the comments.php page, and the puzzle player window appeared. After a minute or so it still had only plain white content, the status information still said "waiting for data from puzzlemix.com". I closed the window that was obviously not being delivered with a puzzle. Back on the comment.php page the Icon next to the puzzle link was still a green "unplayed" arrow (I did not update the page manually). So I guess the player window didn't tell the comments window that the puzzle was started at that point.
Now I'm listed in the right column even though I had only one session...

Related problem: Sometimes my player window is stuck with "waiting for analytics.google" for a couple seconds before(!) showing the puzzle. Still this time counts as if I were able to use it to solve anything.

Is there any way to change that?

Posted 10th Jul 2020 at 16:35The puzzle player doesn't actually load anything else (it just displays that message while it waits for the javascript player to initialise), so if it doesn't go away it means the https fetch timed out while loading the HTML for the page.

It shouldn't wait for the google analytics code and I thought I checked several years ago to make sure it didn't - and the code hasn't changed in several years in that respect. You are possibly just seeing the coincidental effect of your connection pausing completely - so the player has not fully loaded and similarly the analytics script is waiting to be loaded.

The site is checked every minute from several locations around the world (using updown.io) to ensure every response takes less than 2 seconds, so if you are seeing longer delays then is it possible your internet connection is problematic?

Posted 10th Jul 2020 at 16:37(But more generally, it would be better if the player said 'okay, ready' to the server once it was up and running, rather than starting the time on page fetch, I do agree)

Posted 10th Jul 2020 at 21:15of course it has to do with my internet connection, that is the wireless from my computer to my internet router (different building - a repeater failed... digging up the ground and tearing up the walls to put a cable in - that's another project...). Sometimes I have a feeling it's not my solution time but the speed of my internet that is ranked in the listings :-)