The APL Orchard

« first day (2918 days earlier) ← previous day next day → last day (26 days later) »

Adám

05:41

@Silas Oh, right, it should have said that.

7 hours later…

Rosario

12:36

Is matrix upper triangular?

Adám

yes, indeed.

3 hours later…

Stack Exchange Feeds

16:00

Q: Dyalog APL does sum `+/` from the wrong end, violating the specification

Of course, many functions f are not associative in the sense that a f(b f c) (which is often just written a f b f c in APL) is not always the same as (a f b)f c. This is also the case when f is addition + of floating-point numbers. Fortunately, in APL there is a simple convention to evaluate from...

4 hours later…

Paul Mansour

19:34

It appears ⎕VGET and ⎕VSET can take a vector of namespaces as the left argument. Is there a reason it doesn't like a matrix or higher rank array?

Adám

@PaulMansour My original spec allowed any shape, but I wanted to be conservative. We can always add it if called for.

Paul Mansour

I'm calling ;)

Adám

Well, I s'pose it isn't too late. Can show some compelling usecases?

Paul Mansour

Well, what's the use case for a vector?

I have a vectors of namespaces all the time, and matrices, but less often

Adám

A lot of data is stored as JSON, including arrays of objects. Those are all "vectors of namespaces"

Paul Mansour

19:40

It was a rhetorical question.

Adám

(⍴⍤⊣⍴,⍛⎕VGET) :-)

Paul Mansour

In APL we obviously have matrixes, so matrices of namespaces will arise. One example is modeling an HTML table of <td> elements.

Yes, nice and I figured it could be done with just simple selective assignment too...

I don't think it need to go into 20, but it just seems like if you allow a vector, there would be no reason to not allow arrays of any rank.

Adám

@PaulMansour That's a very good example. OK, I'll take it up.

@PaulMansour I really want to be able to write ⍢, one day…

@PaulMansour Well, for ⎕DT, we do allow any rank, so why not here?

Paul Mansour

Is that "under"?

Adám

Yeah.

Paul Mansour

19:52

While I have you, is there a keystroke for getting out of an multiline edit situation in the session in V20?

Without having a syntactically correct thing?

Adám

Yes, issue a (weak) interrupt. Press the Break keyboard combo, usually Ctrl+Pause although Pause may be hidden behind an additional Fn shifting key. On this laptop (Dell), Ctrl+Fn+B does the trick.

Action>Interrupt from the menu and Weak Interrupt from the systray does the same thing.

Finally, you can assign a keyboard combo to the menu item: ⎕SE.mb.action.interrupt.Accelerator←73 2 makes it Ctrl+I

You can find the numeric code for a keypress by running ]keypress, pressing the key and using the codes from elements 5 and 6

Paul Mansour

OK, thanks a bunch. Will definitely set up a key for that (or turn off multiline editing!). I feel like it should be Escape but not sure...

Adám

Escape (as the traditional way to cancel or, well, escape interface states) would make sense if it wasn't for Dyalog's odd decision to make Escape mean save+close.

Paul Mansour

20:09

Yes indeed. But I already use Escape to get out of a trace session, so its not too much of a stretch.

Adám

@PaulMansour ⎕SE.mb.action.interrupt.Accelerator←27 0 then. To my surprise, it seems to work well, without removing existing functionality.

« first day (2918 days earlier) ← previous day next day → last day (26 days later) »

all rooms