MATL CHATL

Meh. It's just making the input suit my needs. And still, I'm not sure if it complies because the OP is super vague about non-unique entries. But a single u could fix that if necessary.

I liked my 14 byte program a lot better, because it exploited the idea of unique entries to save me like 5 bytes.

Luis Mendo

Yes, that challenge is badly specified. I VTC'ed after they failed to clarify

and downvoted, I now see

The challenge is quite unclear

Sanchises

Yeah me too but I couldn't resist tying with Jelly.

(or well, first winning, but then it turned out arbitrary strings had to be supported)

I have no morals.

By the way, this is MATLAB behaviour too, but why would {{1,2},{3,4}}"@&DD yield {{1,2}} and {{3,4}} instead of just {1,2} and {3,4}?

It's wrapped in some extra layer of cell array for some reason.

Maybe I'll just ask in the MATLAB/Octave chatroom

...or on SO, because why not.

Luis Mendo

12:58 PM

@Sanchises It turned out to be a very silly, easy to fix mistake in my patch functions for Octave (a 2 should have been 1). It also affected setdiff, union, intersect. All of them are fixed now. Thanks!

Commited:

* Correction for set functions in Octave. The bug afffected Octave's compatibility functions for setxor, setdiff, unionand intersect. Thanks to Sanchises for noticing!

@Sanchises Yes, that's what Matlab does. I was tempted to change that in MATL, but I decided to go with Matlab, although that behaviour is hardly evver useful

Basiccally, Matlab's for applies () indexing (for the columns of the array), even if the loop is over a cell array

See third item here:

> For example, on the first iteration, index = valArray(:,1). The loop executes a maximum of n times, where n is the number of columns of valArray, given by numel(valArray(1,:)). The input valArray can be of any MATLAB data type, including a character vector, cell array, or struct.

Sanchises

1:40 PM

Thanks for your answer on SO.

Yeah it's better to keep it at MATLAB's way, you don't want sudden bugs popping up when you for example loop over a cell array that can be either a row or a matrix.

You could consider a &" which checks whether it's a row and unwraps it accordingly. Or vice versa, do that by default and use &" for true MATLAB behaviour.

But I don't know if the language fundamentals allow for & to precede a control flow operator.

2 hours later…

Luis Mendo

3:25 PM

@Sanchises It was an interesting question :-)

@Sanchises That's the problem. & (like $ and # ) only applies to normal functions, not to control statements like ". It would have to be a new 2-char control statement. And it's probably not worth it, because almost always the cell array is a vector, so to get the cell contents in each iteration you only need to use g (cell2mat) after @

Sanchises

@LuisMendo Yes, I figured. I'm getting the hang of how you guys built the compiler a bit, and I already assumed that wasn't easily supported.

What actually happens if you do &"...?

Does the & just apply to the first loop, first function, or?

Luis Mendo

3:40 PM

@Sanchises Building it was fun :-) The main problem is that visualizing the function definition file (funDef.txt) may be difficult, for example on GitHub, because of the tabs

@Sanchises & applies to whatever normal function comes next. A normal function is that which is not a loop/branch, or stack-rearranging function, or meta-function, or clipboard function

Example:

>> matl 3:&"@f
1
1
1
1

You get four 1's here because the first f has the & modifier applied, which turns it into two-output find

Sanchises

4:01 PM

Yeah I thought so. Good example, thanks.

Why did you go for a single definition file, and not a large (well... Very large, there may be my answer) number of m files?

Luis Mendo

You may be thinking in terms of Matlab functions. But the MATL compiled file (MATLc.m) doesn't have functions (except functions like setxor that need to be intercepted for Matlab compatibility; but that's a different story). A MATL function like f is translated into a bunch of lines directly in the compiled file.

So what is needed is a way to specify the lines that define each MATL function. Apart from some wrapping code to get values from/onto the stack, implicit input etc, which is common to most functions, those lines actually define that the MATL function does. For example, for f it's just [out{:}] = find(in{:}) (in and out are handled by the wrapper code I mentioned). For other functions there are more lines, but often it's just something like the above: a simple call to a Matlab function

Those lines are defined in the 13th column of funDef.txt. Other columns define things like possible and default numbers of inputs and outputs, help/doc text etc

Sanchises

4:43 PM

@LuisMendo Yeah I see. But there is no real reason to read those functions from one file, or read them from multiple files. For example, you did have separate compatibility function files, rather than make a compFunDef.txt file.

Luis Mendo

Yes, it could have been done differently. Anyway, they are not directly read from the file txt. That file is converted into a funDef.mat file, which contains a struct array with the data. So when a function needs to be inserted into the compiled file, its defining code is taken from the body field of that struct, which corresponds to the column I mentioned from file funDef.txt

3 hours later…

Luis Mendo

7:40 PM

Release 20.4.0:

* Circular convolution (`cconv`) implemented as `Z+` with 3 inputs; including Octave compatibility
* `Zc` with a single, non-cell input replaces nonzeros by 35 (ASCII for '#') and converts to char. Thanks to Conor O'Brien for the idea!
* Octave compatibility: `Zd` (`gcd`) and `Zm` (`lcm`) now allow char inputs
* Extended `Zs` with skewness and kurtosis
* Correction for set functions in Octave. The bug afffected Octave's compatibility functions for `setxor`, `setdiff`, `union` and `intersect`. Thanks to Sanchises for noticing!

@Suever I just created a new release, and it failed to automatically push to MATL Online again. Github says error was HTTP 500 (same as tge other time). Manually redelivering didn't work. Maybe it has become read-only again?