Sort question

[dunbarx]

Richmond.

As I said, I am not smart enough to have made that up. It lives somewhere.

And I cannot imagine, though I will check tomorrow, that it might work in HC but not in LC.

Craig

[FourthWorld]

I can recall seeing some sort of "and" clause for something like this, so if Craig's crazy he and I suffer from the same affliction.

Was it perhaps related to marked cards?

[FourthWorld]

Found it: page 366 of "HyperTalk 2.2: The Book" describes the "mark" command as supporting multiple search options with "and".

The LC Dict has no such example, but attempting it throws no error (I haven't yet put together an example to see if it actually works).

If this does indeed work in LC for using multiple search criteria, it would be helpful to see that part of the parser extended for use with the "sort" command and possibly others as well.

[richmond62]

Aha:

I was looking in 'THE COMPLETE HYPERCARD 2.2 HANDBOOK'.

And, now you have told me what to look for . . .

page 541 in that books explains about mark for selective searches.

[dunbarx]

Richard, Richmond.

This is entirely different. Everybody knows one can

Code: Select all

mark cards where thisIsTrue and thatIsFalse

That is a simple boolean.

What I read somewhere was strictly applied to the sort command, that created its own special compound "sortKey" that belied the on-the-face-of-it concatenation that, in fact, such a line of code seems to very standardly imply.

Whew.

Anyway, thanks for looking so hard. I did not make this up.

Craig

[RCozens]

Craig, et al,

The button script in this stack will give you the result you desire in a single sort.

[dunbarx]

Hi.

The handler you posted last week does this best. No kludges or custom functions, just a little automation exploiting the very powerful and stable sort command we love. One need only determine the number of words in the longest line and set the "repeat with" loop counter to that value.

I actually made a gadget that determined the longest word in a body of text, and appended all shorter words in a line with charToNum(2) so they were all of equal length. Of course, those extra padded char(s) all sorted first. A kludge and a half for sure.

But the stable sort inherent in LC (and in HC before it) makes the whole exercise unnecessary except for those who have nothing better to do than fool around with this sort of thing. Like me.

Craig

[RCozens]

But the stable sort inherent in LC (and in HC before it) makes the whole exercise unnecessary except for those who have nothing better to do than fool around with this sort of thing. Like me.

And moi: after posting my single sort script I realized it is unnecessary to pad the third word because it is the final element of the sort key.

This script produces the same result with four less lines of code and demonstrates how a function can be used to script custom sorting logic :

Code: Select all

local word1Max, word2Max

on mouseUp
   put field "Source Text" into text2Sort
   repeat for each line line2Sort in text2Sort
      put max(length(word 1 of line2Sort), word1Max) into word1Max
      put max(length(word 2 of line2Sort), word2Max) into word2Max
   end repeat
   sort lines of text2Sort by justify(word 1 to 3 of each)
   put text2Sort into field "Sorted Text"
end mouseUp

function justify word1,word2,word3
   repeat while length(word1) < word1Max
      put " " after word1
   end repeat
   repeat while length(word2) < word2Max
      put " " after word2
   end repeat
    return word1&word2&word3
end justify

Cheers!

[rkriesel]

This script produces the same result with four less lines of code ...

Since you're interested in less code, how about this:

Code: Select all

sort tText by pad( word 1 of each ) & pad( word 2 of each ) & word 3 of each

function pad tString
   put "" into item 1000 of tString
   replace comma with space in tString
   return char 1 to 1000 of tString
end pad

— Dick

[bn]

Since you're interested in less code, how about this:

Code: Select all

sort tText by pad( word 1 of each ) & pad( word 2 of each ) & word 3 of each

function pad tString
   put "" into item 1000 of tString
   replace comma with space in tString
   return char 1 to 1000 of tString
end pad

— Dick

And picking up Dick's idea how about this

Code: Select all

sort tText by ( word 1 of each ) && ( word 2 of each) && (word 3 of each) & space

It seems to work but I only have vague ideas about why...
But it does have fewer lines...

test case

Code: Select all

ABC XXX XYZ
AB XXX XYZ
a XXX XYZ
A XXX XYZ

Kind regards
Bernd

[dunbarx]

AHA.

Bernd, my man.

Maybe I misremembered the syntax from 30 years ago. This works:

Code: Select all

sort it by word 1 of each && word 2 of each && word 3 of each

The "&&" instead of "&".

And now I (we?) have to think why, since there is still concatenation, just with a space between the several clauses.

Craig

[dunbarx]

And if, after more testing and discussion, this seems to be reliable, surely a note in the dictionary is appropriate. It is, after all, adorable.

Back in the day, when users had the ability to add their own comments to the dictionary, I did indeed include this ditty in the "sort" command entry. But I do not remember if I used "&&" between clauses or not. I hope so...

Craig

[dunbarx]

Once again, I feel like Richmond with my own string of posts.

So what is going on within the bowels of the "sort" command? This test, with a field 1 containing the alphabet soup we have been using, fooling around to see what comes up:

Code: Select all

on mouseUp
   get fld 1
   put word 1 of each && word 2 of each && word 3 of each into sortGadget
   breakpoint
   sort it by sortGadget
end mouseUp

Just to see what concatenation might turn up, right? Try it. check out the variable "sortgadget". I can see that the whole thing overall ought not to work, actually. But I do not see what is going on.

Craig

[dunbarx]

Me again.

I cannot find a flaw in the new one-liner sort variant. Fooling around with various combinations of characters in all the words in all the lines of a body of text, especially using words of varying lengths within the lines themselves seems to work correctly.

I still wonder where I first read about this. I am (almost) positive it was from Goodman.

Craig

[RCozens]

Morning All,

From what I'm seeing it seems that LC can sort words of different lengths correctly, and that the error initially reported was caused by concatenating the words.

So my one liner is:

Code: Select all

sort lines of text2Sort by word 1 to 3 of each

Still, I learned it is possible to code different sorting logic, which could include factors beside the text itself, by building sort keys via a function.