Post by: Qwerkey on August 06, 2018, 05:32:00 am
The program provides all the solutions to a particular kind of puzzle (Trackword) which appears in a specialist magazine. Many years ago, I originally created a program for this in QuickBasic 2, and I have used this as a basis for learning to use InForm. InForm is so user-friendly that there was only a little work needed to convert from the original text-only program. In fact, the programming difficulty and interest was in the original work in creating the algorithm which finds all the solutions (and only those which are valid). Looking at the code now, it is quite difficult to work out how the algorithm works exactly. I must have been brighter all those years ago.
Should you be interested to run this specialist program, you will need to download the pdf. The solving routine needs tens of thousands of computational cycles and needs a modern fast machine. A slower machine will give a noticeable pause before answers are displayed. I should have programmed a progress bar, but I thought that this little starter project did not justify it.
Richard
[Edit: Manual updated 16/2/19]
Post by: bplus on August 06, 2018, 09:03:53 am
Once again my system is hung by having to remove ../ from include files but after that everything worked great!
33 words found in "Trackword" word.
Very nice use of Inform!
So it's not just the permutations of 9 letters but all the permutations of letters up to 9 and then checking it's a word.
But you probably don't check all permutations (because on my system that takes a while just to do 9 letters) but just find words with same group of letters? Darn quick finds anyway!
Post by: bplus on August 06, 2018, 09:20:40 am
Just check each word to see if it has same letters as in the starting 9.
That algo might be same as bulls and cows (but only checking cows), except you can quit word as soon as there is no matching letter.
Post by: bplus on August 06, 2018, 09:35:22 am
I am going dig up my algo for counting cattle.
EDIT double check BPLUSLOOP no loop, loops or plus, but LOOPBPLUS is good!
Post by: Qwerkey on August 06, 2018, 09:38:57 am
Wait, it could probably be done with one pass through dictionary!
Just check each word to see if it has same letters as in the starting 9.
That algo might be same as bulls and cows (but only checking cows), except you can quit word as soon as there is no matching letter.
The puzzle is specific in that the word has to snake through the grid, so letter order is specific. I am spending some time to speed up the search algorithm, so any additional ideas will be gratefully accepted. The computing time time taken up forming all the 10,256 allowed letter combinations, not so much checking the dictionary.
The ../ which I put (was not in Fellippe's original), I thought was necessary because of the /Trackwork extra folder layer and I thought the ../ wasrequired to get back to the QB64 root folder. Another thing that I have failed to understand. I shall stand in the corner with a pointed hat on.
Richard
Post by: bplus on August 06, 2018, 09:46:56 am
Post by: bplus on August 06, 2018, 10:03:16 am
Quote
I am spending some time to speed up the search algorithm, so any additional ideas will be gratefully accepted.
Hi again,
Steve has a Spellcheck program here:
http://qb64.freeforums.net/thread/33/spellcheck
I will bet it is pretty darn quick!
Do the words have to appear like in WordSearch in the grid, ie backwords or forwards, across, up/down or diagonal?
Post by: bplus on August 06, 2018, 10:11:42 am
EDIT: Oh and the speed bump is NOT looking up the words but running the snakes.
Post by: FellippeHeitor on August 06, 2018, 12:57:54 pm
This is my first program created with InForm. Firstly, I must reiterate members' comments on what a superb achievement is Fellippe's work in creating InForm. I have previously attempted to code in Microsoft's Visual Basic 6 (just about usable with no graphics available) and then their Visual Studio (completely unworkable). Fellippe has achieved everything that we should have desired from the Microsoft stuff with much easier usage and integrated into all that we wish to do in QB64. Hats off again to Fellippe.
Thank you so much, Richard! That makes me want to go even further in the development of InForm, I really appreciate it.
The program provides all the solutions to a particular kind of puzzle (Trackword) which appears in a specialist magazine. Many years ago, I originally created a program for this in QuickBasic 2, and I have used this as a basis for learning to use InForm. InForm is so user-friendly that there was only a little work needed to convert from the original text-only program. In fact, the programming difficulty and interest was in the original work in creating the algorithm which finds all the solutions (and only those which are valid). Looking at the code now, it is quite difficult to work out how the algorithm works exactly. I must have been brighter all those years ago.
Should you be interested to run this specialist program, you will need to download the pdf. The solving routine needs tens of thousands of computational cycles and needs a modern fast machine. A slower machine will give a noticeable pause before answers are displayed. I should have programmed a progress bar, but I thought that this little starter project did not justify it.
Richard
I love it! You seem to have gone through InForm's wiki since you used many methods we didn't even get to discuss over email. Great job there.
If you wanted to make it more portable, you'd need to include 4 extra files in your zip, so that one wanting to run your program but not interested in the full InForm package could still do so:
- InForm/InForm.ui
- InForm/xp.uitheme
- InForm/InFormVersion.bas
- falcon.h
One last thing I forgot to ask: Did you download the zip package for InForm or did you use the installer?
Post by: Aurel on August 06, 2018, 03:06:24 pm
I don't know that qb64 can produce" native" GUI apps.
So i s that wrapped trough sdl libs or is used native wrapper to win32 api calls( CDCL).
Why i babeling about this is that i have written almost complete set of GUI funcs but writtebn in o2
BUT can be build into one dll.
Post by: bplus on August 07, 2018, 01:16:34 am
Quote
The puzzle is specific in that the word has to snake through the grid, so letter order is specific. I am spending some time to speed up the search algorithm, so any additional ideas will be gratefully accepted. The computing time time taken up forming all the 10,256 allowed letter combinations, not so much checking the dictionary.
Hi Qwerkey,
I have worked out a 2 part system:
Part 1 builds a one time only, Master Snakes.txt file, that contains the 10256 letter snakes that can be formed on the 3x3 grid.
The snakes are listed as strings of square numbers. This is done once for all time, no need to do again.
Part 2:
Loads the snake patterns, the dictionary words and gets down to business by translating the square numbers to letters in the given word grid or just a 9 letter word string. With the word formed it checks the word list with binary search and stores all matches in array that is sorted and then displays and recounts only unique words.
You might like to see the little recursive gem that is only about a dozen lines called "build2" that gets all snake patterns for each square of grid, the heart of the whole thing.
Post by: Qwerkey on August 07, 2018, 04:03:35 am
QuoteThe puzzle is specific in that the word has to snake through the grid, so letter order is specific. I am spending some time to speed up the search algorithm, so any additional ideas will be gratefully accepted. The computing time time taken up forming all the 10,256 allowed letter combinations, not so much checking the dictionary.
Hi Qwerkey,
I have worked out a 2 part system:
Part 1 builds a one time only, Master Snakes.txt file, that contains the 10256 letter snakes that can be formed on the 3x3 grid.
The snakes are listed as strings of square numbers. This is done once for all time, no need to do again.
Part 2:
Loads the snake patterns, the dictionary words and gets down to business by translating the square numbers to letters in the given word grid or just a 9 letter word string. With the word formed it checks the word list with binary search and stores all matches in array that is sorted and then displays and recounts only unique words.
You might like to see the little recursive gem that is only about a dozen lines called "build2" that gets all snake patterns for each square of grid, the heart of the whole thing.
bplus, wow you've certainly taken an interest in this. I have just started a method where the 10,256 confiurations are created then used (and maybe on 5,128 and then reverse the word), so I'll certainly look inot what you've done.
Thanks very much, Richard
Post by: Qwerkey on August 07, 2018, 04:08:05 am
If you wanted to make it more portable, you'd need to include 4 extra files in your zip, so that one wanting to run your program but not interested in the full InForm package could still do so:
One last thing I forgot to ask: Did you download the zip package for InForm or did you use the installer?
Fellippe, I'll look into the portability thing. That looks more flexible. I've more reading of the wiki to do - you did a great job with that, as well. I used the Installer from the Website.
Richard
Post by: Qwerkey on August 07, 2018, 05:53:39 am
Existing method (non-Inform, tetxt only): 0.768s
Pre-prepared (as bplus, but my coding): 0.69s
Existing method in Inform: 1.15s
So, firstly changing to InForm, but exactly the same search/check routine slows the process down (not quite sure why). Secondly, there is a slight improvement with the pre-prepared method, but I've not yet optimised this or used bplus exact method.
Richard
Post by: FellippeHeitor on August 07, 2018, 06:20:24 am
Code: QB64: [Select]
Just don't forget to turn those timers back on after you're done:
Code: QB64: [Select]
InForm's events and repaint routines are timer-based and that surely takes up some cycles. The tradeoff of the above method is that your interface will remain unresponsive while the calculations take place. You must judge if the speed gain is considerable enough to justify the tweak.
Post by: Qwerkey on August 07, 2018, 06:35:45 am
Post by: FellippeHeitor on August 07, 2018, 06:41:55 am
Post by: Qwerkey on August 07, 2018, 06:51:00 am
Incredible!
Post by: FellippeHeitor on August 07, 2018, 07:00:21 am
Post by: Qwerkey on August 07, 2018, 07:08:16 am
When I was considering putting in a Progress Bar, I assumed that the solve routine would have to be placed in the SUB __UI_BeforeUpdateDisplay subroutine with a 30 times / second (default) _DISPLAY update. But as this program hasn't left the SUB __UI_Click (id AS LONG) subroutine, it shouldn't be affected by any InForm stuff? You can tell that I haven't got a clue as to how machine-level code (or any other level come to that) works.
Richard
Post by: FellippeHeitor on August 07, 2018, 07:28:25 am
What I'd do if I wanted a ProgressBar control to be updated while the calculations take place would be something like this (pseudo code):
DIM SHARED progressValue AS LONG
DIM SHARED isCalculating AS _BYTE
'in sub Click:
if id = solveButton then
isCalculating = True
maxStep = whateverRichardSaysItIs
control(ProgressBar).max = maxStep
for progressValue = 1 to maxStep
'do your thing
next
isCalculating = False
end if
'in sub BeforeUpdateDisplay:
if isCalculating then
control(ProgressBar).Hidden = False
control(ProgressBar).value = progressValue
else
control(ProgressBar).Hidden = True
end if
Notice how BeforeUpdateDisplay sub will deal with whatever is going on in sub Click.
Post by: Qwerkey on August 07, 2018, 10:57:48 am
When your algorithm 2 is used, the time taken to do the search is 0.66seconds, compared to my values above (it's slightly dependent upon which word is searched). So values are similar. However, that time includes reading the dictionary file into the array, which you have to do.
The time taken to do the seach (after the dictionary is loaded) in your method is 0.05seconds. That is quite a change indeed.
In my program, I do the dictionary search directly from hard disc file. I had assumed that the computation time would be taken up in doing the text manipulations to create the 10,256 words. But it does now seem as if it is the dictionary search which is taking the time. I'll try your method in my code.
Thanks a lot, Richard
Post by: bplus on August 07, 2018, 11:16:02 am
You are most welcome! Thank you for such an interesting post I seemed ready to help.
Post by: Qwerkey on August 08, 2018, 05:30:42 am
Richard
Post by: bplus on August 08, 2018, 08:55:12 am
Make sure you are loading files once at the beginning of program, a one time only event and you definitely want event checking off while doing that along with using Fellippe's ideas for turning off event checking while doing the calculations.
As far as using my actual name, my first name is Mark, my initials are MGA, and that is as far as I am comfortable revealing personal info at moment.
On 2nd thought, the file / array loading code could be done in main section of bas file, before any event? Again check with Fellippe.
PPS Again I remind you, Steve's methods / example used for his spell checker program are probably way better optimized than what I wrote in a night just to test the concepts. Before the binary search method was employed (the found sub), my word find (brute force) was really slow too!
PS ^2 use "QB64 Team" as my last name!
Post by: Qwerkey on August 08, 2018, 12:03:19 pm
I'll definitely do all the loading, array addressing as a one-off at program start. I'm happy to copy exactly what you've done: it works so well. I'd be grateful if you'd just give your search function another check over.
Code: QB64: [Select]
- test& = (hi& - lo&) / 2 + lo&
- lo& = test&
- hi& = test&
- 'if still here then
- found% = 0
My equivalent of this would use test& = (hi& - lo&) \ 2 + lo& instead of test& = (hi& - lo&) / 2 + lo&. And when the RHS is odd, yours gives one integer higher than mine (which rounds down - your rounds up).
I ought to be able to work out which is the correct one, but the (simple) algorithm scrambles my brain. If you're happy that yours always finds any match as it clearly seems to, I'll go with that.
Thanks a lot, Richard
Post by: bplus on August 08, 2018, 01:00:35 pm
Without researching carefully, I would strongly guess it doesn't matter whether it is rounded up or down.
Maybe change
WHILE hi& - lo& > 1
to
WHILE hi& - lo& >= 1
to cover our butts? Whatever way, the conservative answer is to keep test& rounded down for low side and rounded up for hi side, so that test& never jumps past our target word.
OK I will test these ideas... stay tuned, unless you've done your own tests already.
I am glad you like QB64 Team, they should get credit! :)
Post by: FellippeHeitor on August 08, 2018, 01:03:05 pm
Post by: bplus on August 08, 2018, 01:11:15 pm
When I say QB64 Team, I mean all who participate in this process.
Does QB64 Team have another meaning?
Post by: bplus on August 08, 2018, 02:06:12 pm
Code: QB64: [Select]
- _TITLE "Hi Lo Game AI"
- ' Hi Lo AI.bas SmallBASIC 0.12.9 (B+=MGA) 2017-08-14
- DEFLNG A-Z
- secret = 0
- secret = 11 'another limit test
- 'test intelligence of this program with boner! comment next code line
- ' for limit tests comment next line or uncomment for normal range test
- Hi = 11: Lo = 0 '<<<< make these higher and lower than what the secret number can be (just tested secret = 10, oops!)
- guessNumber = 0
- WHILE Hi - Lo > 1 '<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< test both > 1 AND >= 1 Result: Does not work with >= 1
- guess = (Hi - Lo) \ 2 + Lo '<<<<<<<<<<<<<<<<<<<<<<<< test both / AND \ Result: Both work OK
- guessNumber = guessNumber + 1
- PRINT "Yes! The secret number was "; guess
- INPUT "OK, press enter"; dummy$
- PRINT "Too high!"
- Hi = guess
- PRINT "Too Low!"
- Lo = guess
- PRINT "Yikes! Programming error, did not find secret number."
- PRINT "WTH? did someone mess with secret number?"
What could go wrong is NOT setting Hi, high enough at start or Lo, low enough at start. (And I don't think I did make Hi high enough AND Lo low enough with my test code. Fortunately a word wasn't first or last.)
While Hi - Lo > 1 keep testing, NOT While Hi - Lo >= 1, it will never finish!
It did NOT make a difference if / or \ was used.
Sorry, code needed further edits.
Post by: FellippeHeitor on August 08, 2018, 02:19:22 pm
Does QB64 Team have another meaning?
Well, kind of (https://twitter.com/qb64team). Luke, STxAxTIC and I maintain qb64.org under that name, so I guess that's what you meant after all?
Post by: bplus on August 08, 2018, 04:51:23 pm
Does QB64 Team have another meaning?
Well, kind of (https://twitter.com/qb64team). Luke, STxAxTIC and I maintain qb64.org under that name, so I guess that's what you meant after all?
Hi Fellippe,
Well I know you 3 above in quote as Odin from here of course, and I remember you using the "QB64 Team" name with the Christmas Challenge on Facebook (from NET) and assumed you were referencing all the people who submitted code, STxAxTIC and Luke had not participated then with that as I recall.
I am totally unfamiliar with Twitter stuff and link, so I didn't know.
If "QB64 Team" name is "taken", I am sure we can come up with something else that is satisfactory to all.
Maybe "Team QB64"? ;-))
Append:
Oh it was Happy 2017 thing not Christmas:
BTW, only time I spent any time on Facebook was to check that 2017 thing out.
Post by: SMcNeill on August 08, 2018, 06:13:51 pm
Maybe "Team QB64"? ;-))
How about The QBCoders? With a nice QBC logo drawn over by a 64 for a pretty visual.
Post by: bplus on August 08, 2018, 07:12:04 pm
the interaction or cooperation of two or more organizations, substances, or other agents to produce a combined effect greater than the sum of their separate effects.
Nice goal.
Post by: Qwerkey on August 10, 2018, 04:52:38 am
Fellippe, you mentioned making the program portable which is a good idea.
If you wanted to make it more portable, you'd need to include 4 extra files in your zip, so that one wanting to run your program but not interested in the full InForm package could still do so:
InForm/InForm.ui
InForm/xp.uitheme
InForm/InFormVersion.bas
falcon.h
I tried moving the four files into the Trackword folder, and changed
'$INCLUDE:'..\InForm\InForm.ui'
'$INCLUDE:'..\InForm\xp.uitheme'
to
'$INCLUDE:'InForm.ui'
'$INCLUDE:'xp.uitheme'
The IDE was happy, but failed C++ compilation. As you know I like to supply all my work in a folder which can be deleted when no longer required - it makes it easier than trying to remove individual files from the QB64 directory. At the moment, then, I haven't learnt how to make the program independent of InForm.
Richard
Post by: Qwerkey on August 10, 2018, 05:12:37 am
': Trackword Solver by QWERKEY (Richard Notley) 09-08-2018
': bplus created search algorithms
': This program uses
': InForm - GUI library for QB64 - Beta version 7
': Fellippe Heitor, 2016-2018 - fellippe@qb64.org - @fellippeheitor
': https://github.com/FellippeHeitor/InForm
The first line is untrammelled vanity by Qwerkey.
The second maintains bplus's privacy.
The next lines come automatically from the InForm compilation, and rightly acknowledge Fellippe's magnificent work.
We acknowledge the QB64 Team (Team QB64?) by default.
Incidentally, I'm slightly mystified by the Team names "Odin" (I am picturing a stereotypic wise old Nordic god) and "The Librarian" (I am picturing a stereotypic bespectacled quinquagenarian spinster)?
Richard
Post by: bplus on August 10, 2018, 09:37:19 am
There are 4 files to be included in the InForm project folder (and YES! A folder is best way to distribute.)
These are over and above the minimal two you develop for your particular app.
Your folder still needs the falcon.h and the InForm.bas file you created your InForm .bas and .frm files with.
Fellippe called that one InFormVersion.bas (not to be confused with the .bas file you create).
I think the falcon.h file might have to be stored in same folder as the QB64.exe version you are using. It is hard to tell because that is where mine is. I suppose I could test by renaming and seeing if the falcon.h file works from the project folder.
These 4 files are needed to reconstruct and compile the .exe from QB64 so that the exe is for the users platform.
Also the Binary search routine:
When I tested for differences between / and \ ( no difference) I did discover I did not set the Lo low enough and the Hi high enough at the start of the search, did you catch that when rewriting the code? If you missed it the first and last words from the dictionary will not be "found" by the function of that name and it will return a false false (a 0 when it should return 1 for found).
Post by: bplus on August 10, 2018, 10:19:23 am
On my system I had a folder with the falcon.h file on same level as .bas for app and also in the Inform subfolder of the project.
I renamed my falcon.h in my QB64.exe folder falconX.h and the app would not run, but would run when I renamed the falcon.h in folder with QB64.exe.
So those instructions might also be included in the project folder distribution: "Move (or copy) the falcon.h file to the same folder as QB64.exe if you don't already have one there."
Post by: Qwerkey on August 11, 2018, 05:46:00 am
The bplus search method gives a much improved search time, now about 0.1s. This compares to 0.05s which the full bplus method gives. My sort method for the found words slows it down a little and Inform increases the time by a tiny fraction. Even on a slow machine of mine it's pretty instantaneous. We are there!
I thought that I might try to understand the binary search method in detail, and so I went to Wikipedia's expert where the following is given:
Procedure
Given an array A of n elements with values or records A0, A1, ..., An-1, sorted such that A0 = A1 = ... = An-1, and target value T, the following subroutine uses binary search to find the index of T in A.
1. Set L to 0 and R to n - 1.
2. If L > R, the search terminates as unsuccessful.
3. Set m (the position of the middle element) to the floor, [the greatest integer less than or equal to], (L + R)?/?2.
4. If Am < T, set L to m + 1 and go to step 2.
5. If Am > T, set R to m - 1 and go to step 2.
6. Now Am = T, the search is done; return m.
This iterative procedure keeps track of the search boundaries with the two variables. The procedure may be expressed in pseudocode as follows, where the variable names and types remain the same as above, floor is the floor function, and unsuccessful refers to a specific variable that conveys the failure of the search.
Our dictionary array has elements from 1 to 155237 (rather than from "0 to n-1"), so just add 1 to everything.
Then, the following routine (which I have always used in the past) follows this procedure exactly:
Code: QB64: [Select]
- Located` = False: P0& = 1: P100& = Entries&
- Located` = True
- P0& = P50& + 1
- P100& = P50& - 1
You can see that it works by thinking through an array with 1 to 4 elements and then with 1 to 5 elements. So, I have used this method in the search function instead of bplus's. In the actual Trackword program, the BIT variable Located` is replaced by a BYTE variable Located%% as BIT variables are not successfully transferred to/from Functions or Subroutines.
I constructed a program which created an ordered dictionary file with elements 1 to 150,000 and then checked that every element is correctly found and that any word between every element is not found and that a word before and a word after the dictionary elements is not found.
The Wikipedia instruction is that rounding down must be used (as does this function), but I think that because of the P0& = P50& + 1 and P100& = P50& - 1 statements, rounding up also probably works. Anyway, I'm confident that we've done it correctly.
The simplest procedure for me will be to replace the User Manual in the first section of this post with my updated version, but there is no "Modify" procedure on this website. Is there a way to modify what's already there? If I've missed the obvious in usage of the site, I apologise.
Richard
Post by: bplus on August 11, 2018, 09:44:17 am
Code: QB64: [Select]
- The final version is now ready,...
Remember you don't need all of the 4 files Fellippe listed but someone who doesn't have InForm does need all 4 files and to move the Falcon.h to same folder as QB64.exe as mentioned before but doesn't hurt to repeat.
Are you still using disk file access for the Binary search?
Just post the code, forget about modifying a previous post. (Can you tell I am eager to test code?)
Post by: bplus on August 11, 2018, 11:01:18 am
Located` = False: P0& = 1: P100& = Entries& '<<< this starts with Lo one step higher and Hi one step lower
WHILE P0& <= P100& AND NOT Located`
P50& = INT((P0& + P100&) / 2) '<<< this is one less calculation than what I had and I wish now I had started with average instead of average of difference! (It would have saved me from the f... comment!
GET #4, P50& '<<< this, disk file access I still think is slower but are you using that in TrackWord?
InWord$ = RTRIM$(Lex$)
IF Tt$ = InWord$ THEN
Located` = True
ELSEIF Tt$ > InWord$ THEN
P0& = P50& + 1 '<<< ah! a step closer and why not!?
ELSE
P100& = P50& - 1 '<<< ah! also a step closer and why not!?
END IF
WEND
Plus all the variables are typed better.
Append:
An improvement on this code might be to exit the sub as soon as the word is matched and the function return value is set.
With that done the line
WHILE P0& <= P100& AND NOT Located`
can be reduced to just:
WHILE P0& <= P100&
Which is one less logical calculation check and one less variable needed: Located`
Post by: Qwerkey on August 11, 2018, 11:55:50 am
Of course this program uses dictionary searching within an array and not from the disc file (pardon me for the confusion). So, the program load time is now a little longer, but not really noticeable. However, I still use my old method for sorting the found words (which is the part that slows it down a little).
I got the search code from some "standard files" that came with Quickbasic 2.0. Not mine at all.
Richard
Post by: bplus on August 11, 2018, 12:02:21 pm
Append (why start another reply?):
Code: QB64: [Select]
This may explain why EXIT FUNCTION was not used and the flag, Located`, was used.
Post by: bplus on August 11, 2018, 12:30:54 pm
I see a surprise! The number snakes cut in half, you are using word reverse for the symmetric snake that has letters reversed, so from the Master Snakes.txt file you removed all the symmetric reverse snakes. Nice touch!
Now here is the million dollar question (well 25 cents at least) what 9 letter word gives us the most words made from snakes of 3x3 grid? I bet this has been found already, if I know my word enthusiasts.
Post by: bplus on August 11, 2018, 05:54:52 pm
Post by: bplus on August 11, 2018, 11:06:34 pm
Now I am wondering how many are anagrams of each other.
Post by: bplus on August 12, 2018, 01:19:44 am
Code: QB64: [Select]
- _TITLE "Trackword 200 Anagram Analysis"
- 'PRINT AnagramTF%("artss", "stars") 'yeah!
- wTop = 0
- wTop = wTop + 1
- CLOSE #1
- ai = ai + 1
- a$(ai) = w$(i)
- a$(ai) = a$(ai) + " " + w$(j) 'move to i word group
- w$(j) = "" ' this word is used
- 'show results
- l = i - 33: m = 40
- l = i: m = 1
- 'this only works for words 9 repeated letters or less
- 'still here
- aw2$ = aw1$
EDIT: fix with Ltrim$ the str$ of values
Dang! missing last word.
EDIT 2: fixed
Post by: Qwerkey on August 12, 2018, 04:34:19 am
Richard
Post by: Qwerkey on August 12, 2018, 06:01:59 am
If you use the following grid, you gey both CARTHORSE and ORCHESTRA (well-known anagrams of each other):
CAR
RHT
OSE
I think that I remmber Clive Doig (he who invented Trackword) setting a puzzle where there were three nine-letter anagrams. Can there be more than one such solution? I don't suppose that there can be more than three?
Richard
Post by: Qwerkey on August 12, 2018, 06:54:50 am
CAR OSE CRO ETR
RHT RHT AHS SHA
OSE CAR RTE ORC
Working out symmetry considerations is beyond me. Can we reduce HalfABC further by symmetry considerations? Can we reduce search time even more?
Richard
Post by: bplus on August 12, 2018, 09:27:09 am
With the 3x3 grid, there were 3 types of square, the corner, the side, and the middle. The middle was unique but the corner and side had 3 equivalent rotations that likely could be "translated" number for number for the snake patterns and in same manner for the letter patterns. The snakes could be worked out with results from one corner, side and middle but snakes are one time shot. Doing the same with letter patterns would likely take longer to calculate and translate than just translating a snake pattern to a letter pattern, So is it worth it to reduce snake patterns but require and do the same amount of translations?
The reverse trick was quite clever, easiest to understand and use.
Have you any thought about 4x4 grids? Probably learn even fancier words. :D If one bothers to check them out eg, wadsetter :) Which reminds me, do I have an actual dictionary with word AND definition? Such would be a natural addition if you'd like to try Menus with InForm. Heck, it might even be educational, learn while playing... :)
PS a message box would be handy for a quick definition.
Post by: Qwerkey on August 13, 2018, 12:51:56 pm
Richard
Post by: bplus on August 13, 2018, 01:39:12 pm
You have me looking for, and then at, forum stats. Interesting...
Well as Fellippe has mentioned numbers don't exactly mean quality but as I said, quantity is one quality of many...
How does Odin have so many posts? I looked at a quality Odin's posts, only 2 pages not exactly 500+ ???
Append: OK so it doesn't look like a blatant inflation of replies, something more on topic and that I am curious about.
What do you all think of adding an additional capability to define words? To me, it is a most logical, practical, educational and maybe even fun extension of this application of InForm.
Post by: FellippeHeitor on August 13, 2018, 01:41:44 pm
Post by: bplus on August 13, 2018, 02:12:41 pm
Odin wanted to show from the start that post racing would be frowned upon
Post by: Qwerkey on August 14, 2018, 01:19:07 pm
Post by: codeguy on September 04, 2018, 04:55:18 pm
Code: [Select]
WIDTH 80, 43
DIM parray(0 TO 8) AS LONG
FOR i = 0 TO UBOUND(parray)
parray(i) = i
NEXT
PRINT
PRINT
PRINT "permutations"
Permute parray(), 0, UBOUND(parray), np
DO
x$ = INKEY$
LOOP UNTIL x$ > ""
SYSTEM
SUB DisplayResults (PArray() AS LONG, start, finish, np AS DOUBLE)
DIM i AS LONG
PRINT USING "#,###,###,###,###"; np;
FOR i = LBOUND(parray) TO UBOUND(parray)
PRINT PArray(i);
NEXT
PRINT
END SUB
SUB Rotate (parray() AS LONG, Start AS LONG, finish AS LONG)
DIM ts AS LONG
ts = parray(Start)
FOR i = Start TO finish - 1
SWAP parray(i), parray(i + 1)
NEXT
parray(finish) = ts
END SUB
SUB Permute (parray() AS LONG, start AS LONG, finish AS LONG, np AS DOUBLE)
np = np + 1
DisplayResults parray(), LBOUND(parray), UBOUND(parray), np
IF start < finish THEN
DIM i AS LONG
DIM j AS LONG
FOR i = finish - 1 TO start STEP -1
FOR j = i + 1 TO finish
SWAP parray(i), parray(j)
Permute parray(), i + 1, finish, np
NEXT
Rotate parray(), i, finish
NEXT
END IF
END SUB
Code: QB64: [Select]
- '* essentially finds the intersection of 2 arrays
- '* both arrays must be sorted exactly the same order.
- a& = 0
- b& = 0
- b& = b& + 1
- a& = a& + 1
- '* it's a word in both lists
- '* advance both pointers
- a& = a& + 1
- b& = b& + 1
Post by: Qwerkey on September 05, 2018, 06:44:14 am
Richard