@mildused The Wolfram Programming Cloud is free for everyone. If you deploy something it costs Wolfram Credits. Free users get 500 Wolfram Credits per month, and with MMA10 you get an additional 1000 Wolfram Credits I think.
i mean if you think about it: if you have mathematica why settle for 30 minute limits anyways
Yesterday I ran a code that went for 8 ish hours
actually, no it was 11 hours
Let me tell you my story:
I am a 15 year old teenager. I am taking the 3rd Level in a Science Research Course that uses Mathematica to build students up.
I LOVEEE math
I do math competitions every year at National and State Level and place in Top Many times
and whenever I had trouble with Mathematica late night, I wasn't able to call my professor because he was asleep but I can always find a similar problem on Stack and make some similar edits to my code as well
@Rojo Because there are better ways and it's easy to break it. Suppose fun[] does it. Then you (or someone) defines g[...] := ... fun[...] ... that returns a number. Then you use Reap[fun[g[1]], tag] thinking that you'll only get additional results from the fun that you wrote, not knowing or noticing that there's an extra one called inside g. Leaving unreaped sown values is not a good idea IMO.
@acl Well, the rhs also don't get reevaluated once the associations get created
but, the one that I find most troubling for now (but I haven't used them), and is probably a design thing, is that the pattern matching doesn't get inside
So, for example, <|"a"->"b"|> /. {"a"->8, "b"->34}
@Rojo I just meant that that TreeForm is unexpected if you are used to eg Lists, and Part doesn't work like you'd expect from FullForm given how that works for other structures
You have atoms, expressions, and atoms that dress up as expressions. For the latter, half of mma functions fall for the disguise and the other half don't
@acl It was probably too hard to do a PackedArray equivalent for each data structure that needed an internal optimization
I think the pattern matcher, just like AtomQ, hasn't been fooled yet, hummm
Not true
Image fools it to some extent
And association fools most built ins in a peculiar way, that is different from its full form
For example, even though the full form is Assoc[List[Rule[key, val]]], for Level or Depth there's only one level involved. They just treat it as a list indexed by key
instead of integer position
What I meant with "the rhs remains unevaluated" is that ass = <|"a" -> a|>; a = 4; ass
@Rojo That's true. But sw doesn't have to depend on a. It probably doesn't in the OP's real code. It was convenient because a happened also to be the list of positions in a.
By the way, if I do sw = Developer`FromPackedArray[sw];, the first swap takes 0.4 sec (i.e. it's long/slow).
@MichaelE2 Yes, but that unpacks also at level 1, it seems to make a difference
@MichaelE2 Right. And I wouldn't worry much about such internals as long as mma just copies "once when needed", for some definition of needed. But trying to create a copy of sw (not of a) and failing made me start thinking if there are situations where this can be real trouble.
@Rojo Since you are on Linux now too, can you check whether V10 (in the final version, not one of the betas) has weird behavior for 3D Graphics? Everytime I click to rotate, it somehow zooms for a fraction of a second:
@Rojo Additionally, the Antialiasing preference has absolutely no effect. Here with full antialiasing in V10:
When I have joined Mathematica.SE, I unfortunatly logged in into two computers, and couldnt remember my other username, so I created two of them. This one and this.
Is there any way to combine these two users?
Thank you!
@Rojo In CompiledFunction, you can change any specified elements in place without copying the entire array. So I guess there are ways in mma to achieve this, just "undocumented".
@Rojo Ah, since you're here, can you check this behavior? According to the doc "If it is evaluated not within a scheduled task, $ScheduledTask gives None. ", but directly evaluate $ScheduledTask doesn't give None here.
@MichaelHale The problem with this is that Association is atomic, so the minute you make the change to <|a -> b, c -> {{d -> e, f -> g}, {h -> i, j -> k}}|>, you lose the ability to "inspect" further and make replacements. The way out would be to build the associations from inside out.
@MichaelHale Try this:
Block[{$Association},
Replace[json, r : {__Rule} :> $Association@r, {0, Infinity}] /. $Association -> Association
]
@rm-rf @Rojo In rm-rf's first version, why did the /. replace the inner $Associations when the doc for /. says "no further rules are tried on that part, or on any of its subparts."
@Rojo I wouldn't have known to even try that. After I saw ReplaceAll wouldn't work, I figured I needed to use MapAll or ReplacePart with some extra checks.
I am calculating the eigenvalues of a very large matrix, but the calculation of the elements is very slow in mathematica. I want to off load the computation of the elements to C code, so how can I do that.
It should be asked somewhere, can anyone p[oint me with the right link?
@hwlau If it just fits you memory, it is very unlikely you find a procedure that solves your problem without taking any additional memory (for instance for storing the result if it cannot be done inplace).
When I have joined Mathematica.SE, I unfortunatly logged in into two computers, and couldnt remember my other username, so I created two of them. This one and this.
Is there any way to combine these two users?
Thank you!
