Thank you for all the hard work, Andrey - 0.98 stable is a major leap forward from 0.97.

I very much enjoy the encreased computation and GUI performance.

Here is a peculiarity I have found in the recent release:

If I click on the matrix defining expression there is a lag. There is even larger lag when I position the cursor right after "2" in "2000" and hit backspace. In fact it seems (based on the duration of the lag) that expression is fully recalculated when I hit that backspace button - I would think it is not the intent.

Also the resulting size of the uncompressed .SM file (sreenshot attached) is 141 MB (it might have always been this large for large matricies). When I try to copy/paste the expression from attached workbook to a blank one, the destination SMath hangs and is not responsive for good 5 minutes now.

Overall though, those are minor hiccups for the performance boost we ar getting with the 0.98 SMath Studio. Thank you very much

P.S.: one day we will be able to define & manipulate 5000 x 5000 matricies quickly and with ease
matrixSL.smz (1,606kb) downloaded 50 time(s).


This happens on Copy / Paste:

Edited by user 29 March 2016 22:36:27(UTC)  | Reason: Not specified

Thank you Andrey


I see two small issues here:
- CTRL+DRAG doesn't propagate the settings of the new feature (copy+paste works fine);
- no history steps triggered for resizing the matrix;


Nice

Can we have also an update in the SS core documentation?

Davide,


Thank you! Fixed.


Sure, will upload updated documentations after release.

I haven't Matlab C++ Math Library; however, other functions work similarily. And your example looks strange (eval() inconsistently positioned).

This works for me (using Diag() as mwround() replacement):



Here is the cause of your error:



Seeems that without eval(), the vectorize() treats output from other functions (stack in this case) as scalar (it doesn't see the array here before function evaluation - you didn't pass one explicitly), and thus ultimately simply passes its result (that happens to be vector) to the exponent.

Edited by user 30 March 2016 04:59:54(UTC)  | Reason: Not specified

Sorry, I was tired tonight and posted obviously wrong example.
The actual code that puzzled me was:

1-st disabled instance of hb_fir() was certainly working in 5928 but doesn't work any more in 5932.
2-nd disabled also doesn't work
3-rd disabled doesn't work
4-th enabled works fine.



I found that almost in all cases using mwxxx() routines almost as convenient as vectorisation and far more safe.
I'm going to use vectorise() only for component-wise multiplication and division of vectors, to eliminate bulky "for" loops.
And those multiple eval() everywhere are annoying. It is good idea as well to change stack to Mathcad style.
stack(0,v,0) - doesn't work (Mathcad style). Now you have to use eval(stack(stack(0),v,stack(0))) to be sure.

Regards,
Igor

Edited by user 30 March 2016 17:44:25(UTC)  | Reason: Not specified

You can use units for this:

Interesting; using the same approach:



The Jean solution is elegant too.

But in the end I'd prefer the built-in way; note that the issue is related only to the first element in stack (in augment too) -> imho if the first argument of these functions is a number should be considered like a 1x1 matrix (meaning that the desired output will be a column vector or a row vector; at first glance I don't see drawbacks);

Edited by user 31 March 2016 21:07:48(UTC)  | Reason: Not specified