« first day (2062 days earlier)      last day (2723 days later) » 

R. M.
R. M.
01:31
@xslittlegrass Lots of good viewpoints here: mathematica.stackexchange.com/q/86058/5
Ultimately, it mostly boils down to what you specifically need. If your work is computer algebra heavy, then Mathematica is by far the best option. For most other things, you can do it in python with mostly similar performance and you have the advantage that your code is portable and you don't have to worry about licenses, etc.
LegionMammal978
LegionMammal978
Hmm
R. M.
R. M.
Personally, I've never cared for the curated data sources or to format my notebooks using fancy 2D typesetting or using math characters in my expressions, so I don't miss those aspects of Mathematica. Most people won't need like 95% of the shiny features in Mathematica or can live without them.

Some more interesting thoughts here and I agree with the author that inconsistency between packages is the main drawback with python. However in practice, you get used to it (and you can find inconsistencies in Mathematica as well). https://www.johndcook.com/blog/2010/07/09/replacing-mathematica-wit…
LegionMammal978
LegionMammal978
Is there any way to include units in the axis ticks for Plots?
C. E.
C. E.
SameQ's documentation states: "SameQ requires exact correspondence between expressions, except that it still considers Real numbers equal if they differ in their last binary digit."
How can I make such a comparison in, say, the C language?
LegionMammal978
LegionMammal978
@C.E. Something like ((*(int*)&f1)>>1)==((*(int*)&f2)>>1) (maybe with less parentheses)?
(replace with long long for doubles)
(maybe add unsigned as a good measure)
C. E.
C. E.
01:42
@LegionMammal978 Nice, I am actually doing this on doubles, so what is it I need to replace with long long?
LegionMammal978
LegionMammal978
@C.E. So like ((*(unsigned long long*)&d)>>1)
It will take every bit of the double's binary representation except for the last one
C. E.
C. E.
@LegionMammal978 Thank you :)
b3m2a1
b3m2a1
02:09
@xslittlegrass as someone who uses both Mathematica and python (although of late much more Mathematica than python) for recreational work I think python is a smoother (but less fully featured) language to use outside of integrating libraries. It also seems like Jupyter has the Mathematica notebook feel, so give that a look if you haven't. The language is very easy to learn in any case and its user base is growing.
J. M.
J. M.
02:34
@R.M. I would actually even dispute the computer algebra bit. As I'm sure @Nasser knows, there are points where Maple has an edge over Mathematica. The recent diffused attention is not helping improve things in that regard.
xslittlegrass
xslittlegrass
03:19
@R.M. @b3m2a1 Thanks for the informative advice!
 
2 hours later…
J. M.
J. M.
05:02
@LegionMammal978 Have you seen this?
 
5 hours later…
Szabolcs
Szabolcs
09:37
@R.M. How about graphics and visualization? Whenever I try a new system, I tend to like the computational capabilities, but I find plotting such a pain ... No doubt, this is in part because non-trivial plotting is just complicated in any language and takes time to learn. But with Python, I had the impression that it is objectively messier than Mathematica ...
Of course there are a few glaring shortcomings of Mathematica graphics, such as the lack of interactivity (zoom into a plot) or more importantly: creating plot grids.
Szabolcs
Szabolcs
09:48
I really really wish Mathematica had a Python link ...
Then we could use the interesting parts of Python (computational capabilities) without being forced to immediately learn the boring parts in detail (such as plotting).
With the introduction of ExternalEvaluate in 11.2 my hope that M will ever get a decent Python integration is gone ...
2
I frankly do not see people using ExternalEvaluate for anything serious. It is too limited and too slow. It's all about piecing together strings to build Python code, evaluating it in Python, then parsing the output as a string to convert it back to Mathematica.
A useful integration would have fast bidirectional data structure translation, it would support data structures which are critical for scientific computing (numpy arrays!) and it would make it easy to create Python funcions that are easy to call from Mathematica.
LegionMammal978
LegionMammal978
10:41
@J.M. facepalm Thanks, but how would I make this work with the x-axis as well on a log-linear plot?
R. M.
R. M.
11:15
@Szabolcs I miss Mathematica's defaults and its basic functionality for quick visualization (esp. some easy interactive stuff + manipulate) but once you get down to creating publication quality plots, it's pretty much exactly the same in both systems. I've had to write lots of loc of Mathematica + SciDraw code to do things exactly the way I want it and at that point it looks no different than python or MATLAB... it's all a bunch of pixel orchestrations on subplots/axes.
Having your plot be an expression that can be manipulated further w/o redoing the expensive bits in Mathematica is also a nice bonus. However, these are all things that I used to think that I couldn't live without them but in reality, I only miss them like 1% of the time.
LegionMammal978
LegionMammal978
@Szabolcs Why not? Was literally just trying to make an aligned GraphicsColumn of two plots, and they aligned just fine if I manually set their ImagePadding.
halirutan
halirutan
11:50
@Szabolcs A minimal example looks like this. Ideally, we want to evaluate the code from a normal Julia file, precompile it during the library initialization, store the function pointers and create the wrapper Mathematica library functions dynamically with the help of Julias introspection.
J. M.
J. M.
@LegionMammal978 Looks like a good question for main.
halirutan
halirutan
@Szabolcs One weird thing is, that the library crashes when I do the jl_init() step in the WolframLibrary_initialize function. I have no explanation for this behavior because as I understood it, this wolfram function is exactly meant for something like this.
LegionMammal978
LegionMammal978
@J.M. Nah, I just gave up and put the unit in the frame label for that axis
Szabolcs
Szabolcs
12:30
Question to people who have 11.2: Does the new ExternalEvaluate framework even allow to send an array to Python?
halirutan
halirutan
12:57
1 hour ago, by halirutan
@Szabolcs One weird thing is, that the library crashes when I do the jl_init() step in the WolframLibrary_initialize function. I have no explanation for this behavior because as I understood it, this wolfram function is exactly meant for something like this.
Oh my god, how stupid. Never forget to add EXTERN_C even to the initialization function.
LegionMammal978
LegionMammal978
13:29
TIL you should always use "" instead of Null to not put any tick label
Turns out, Null, although not rendering, adds a blank space which causes stuff to look weird
Took me an hour to figure that one out
LegionMammal978
LegionMammal978
14:06
Question: Is there any way to change the font color of a Quantity? Style doesn't seem to be doing much
nvm, TraditionalForm fixed it
 
2 hours later…
J. M.
J. M.
15:44
Anybody here with access to MATLAB R2017a?
murray
murray
16:02
@J.M. With the last version 11.2 Prerelease, I'm seeing the same Tooltip bug discussed in mathematica.stackexchange.com/questions/155434/…. (Didn't want to post that, though.)
J. M.
J. M.
16:23
@murray Well, somebody can put that in as soon as 11.2 is officially released.
b3m2a1
b3m2a1
16:49
@Szabolcs see this. I don't have 11.2 so I can't give you any more info, but you could test it in the cloud.
From this it looks like they want us to use string templates...
I made a partial symbolic python converter to work with psi4 a few months back, that might actually be helpful here. Thing lives here.
 
2 hours later…
Szabolcs
Szabolcs
18:45
@b3m2a1 Yes, I know that, but is it really reasonable to transfer a 100 by 100 matrix as explicitly written code?
The performance will be useless.
My point is that the design is flawed from the beginning. Stitching together code from strings and parsing the output as string is just not usable for much.
@J.M. Yes. You can email me.
b3m2a1
b3m2a1
19:04
@Szabolcs I agree 100%. If the array or other data-structure is large and can't be easily built in python or imported efficiently from a temp file, there's a big performance issue. This is really bare-minimum support.
b3m2a1
b3m2a1
19:24
@Szabolcs so their implementation of `ExternalEvaluate` turns out to be just a simple REPL that probably took ~30 minutes to write. See for instance:

CloudEvaluate[
FileNames["*.py", PacletFind["External*"][[1]]["Location"],
Infinity]
]
Szabolcs
Szabolcs
19:42
I absolutely wouldn't say it took 30 min to write. I did not mean to criticize the work of the person who made it, or the quality of the implementation.

What I did mean to criticize is the design: a one size fits all solution to communicate with any other system will always be slow and limited.
 
3 hours later…
Vitaliy Kaurov
Vitaliy Kaurov
22:55
Why Were Northern Lights Visible from Champaign, IL
http://community.wolfram.com/groups/-/m/t/1179868
user image
user image

« first day (2062 days earlier)      last day (2723 days later) »