Hey y'all. What's the latest version of Workbench? I've got 10.1.822 and just ran into a problem with the "preview documentation" button. Also recently upgraded to 11.3 -- coincidence?

@ChrisK Presumably you mean the current version of the Eclipse plugin?

(I have no idea as I use neither, but just wanted to confirm)

A bug report a day... keeps Wolfram Research support alive?

Curiously RandomSample[ ;; 10^20, 2] fails although there's no real reason why it should.

(I surely send feedback to WRI on these things.)

@kirma That's just RandomInteger[{1, 10^20}, 2] right? I figure they just thought that wasn't an interesting enough use case to special-case in.

@b3m2a1 Not exactly. It avoids duplicates (although one would be pretty unlikely in this case).

Shorter span, say 1 ;; 10^18 does work.

@kirma Oh interesting. Definitely a bug then.

Annoying that things break arbitrarily like that

Support case IDs have grown by 1000 in 12 hours. Frankly I don't even remember what I reported yesterday evening, the automatic response emails could really include the original submission.

@b3m2a1 I'm reasonably confident it's related to span values not fitting into a 64-bit integer, but there's no particular reason why it should fail like that.

Eh, no. Gmail is acting strangely. 1000 IDs in 3.5 days.

Anyone know how to debug a weird 404 error? I'm getting a resource-doesn't-exist issue with the cloud here and on all of my other docs pages which is for reasons unknown attaching itself to the top of the page. This didn't exist when I deployed these.

It says you don't have permission if you aren't logged into the account but really the issue is that something doesn't exist.

Opening the console in Chrome I can't find any explicit "couldn't load ..." errors.

So the cloud isn't reporting the standard 404 message and then it's forcing the page it would have loaded into my page.

I also think this is happening at load-time since this isn't in the raw HTML

And since the page title isn't set to "Page Not Found" for ~.5s after the page is reloaded.

@b3m2a1 your link works for me

You don't get anything like this?

Nope, I get a ref page for create spell index

Same via incognito mode

Well it's good to know that it exhibits the correct behavior for some people at least. I get the ref page following that header.

@Kuba do you see anything weird if you go here: wolframcloud.com/objects/b3m2a1.docs/main.html

?

Bunch of 403

Ugh

The cloud

I admire your persistence to customize cloud in a free time.

I have no idea where those 403s are coming from. I'd assumed it was the same thing causing the 404s (or 403s if not logged in) on the subpages.

I got so annoyed that I only do what I need for work or what is 'given'.

Well I figured I could just use it like a basic webhost and just use entirely static resources but then it breaks without warning like this...

If I got to a separate tab for the first iFrame

wolframcloud.com/objects/b3m2a1.docs/reference/guide/…

I get i.stack.imgur.com/8AmYX.png

@b3m2a1 so it is there

kind of

Yeah I'd worked that out but looking at the actual raw HTML I can't figure out where those 403s are coming from.

me neither

:/

@Kuba found it. It seems Chrome is rendering things like this: <!--#include virtual="https://www.wolframcloud.com/objects/b3m2a1.docs/reference/includes/car‌​ousel-head.html"-->

And that's not in the default resource set package in with DocumentationBuild...

mhm

Oh nevermind it's worse. Even if I deploy a resource there it still links through as a 403...

Ah shit this is gonna suck to fix. This is when I wish I knew a bit more about web dev. Apparently that's rendered on the server side so it doesn't matter that I passed a proper link. WRI would need to handle it properly and I don't trust them to.

This was the issue, by the way: en.wikipedia.org/wiki/Server_Side_Includes

@b3m2a1 you can embedd code to drop those tags

@Kuba it's okay I just added another post-processing rule. I had to fix a bunch of the links that came out of DocumentationBuild anyway. I'll probably just leave most of the subpages with the possible 403s for now as it takes a long time to build out the HTML.

I just patched the main page by hand

Does anybody here have experience on running Mma on Amazon EC2? If I just create a Linux installation, install it there and occassionally start and stop that image on different instance types, do I run into trouble with license management? (Last time I asked from Wolfram their answers were a bit inconclusive, although they didn't really indicate any trouble with my goal.)

@Szabolcs I am having a slightly different take on the issue: As of now, I cannot see good reasons for not retracting close votes - yet they are kept in place. That integrations in 11.2 may have been wrong, that 11.3 does not give answers where 11.2 had given answers and that there is a public WL package called Rubi to solve indefinite integrals

to me seems a) rather relevant for many users (case in point: upvotes), b) calling out for intelligent answers with regard to why this is happening (case in point: Michael E2's answer), and c) has a good practical answer (Rubi) that is useful also for performance reasons (my post and comparing integration times in 11.2 and 11.3). What better fit for M.Se?

I am not too big a fan of Wolfram Community and indeed you have given a good reason yourself: it is not free - I have to log in with my Wolfram ID. At least SE is a "third party". While "not posting bug reports to WRI" may be a viable reason against allowing for bug reports here, I would suggest to have both. And while nobody has any control over what happens with reports, here it is clearly visible.

In most cases I find bug reports / issues a very useful and viable format for M.SE. I doubt that SE would hinder us having them here - and there is chat rooms here also. My understanding is simply that the knowledgebase of Q/A should not be spoiled by discussions and they usually are not.

Last, but not least. In most cases, bugs are in their very essence "Q&A": Is it false behavior or am I to dumb to read the docs? :) Is something else concerned given this issue? What can I do to still get my results (e.g. workarounds). So again: Let us have both: WRI-case reports and Transparency by having them here in the KB.

@b3m2a1 Yes, version 10.1.822 of the Eclipse plugin and 11.3 of Mathematica. I'll poke around with it more this morning to see if I can figure out my problem

@Szabolcs @gwr I don't mind se-bug-reports as long as they are phrased according to se rules. If OP does not want to, anyone can send a report to wri. Most often I don't care if it will be fixed because I need a solution/workaround immediately. Quite often I need to support older versions too.

So I and I guess many users do not care what WRI thinks about specific 'feature' because the problem is they (I) need stuff done and there is an obstacle that should not bet there.

@kirma That's not new, it fails at least as far back as 11.0.

well, at least on mac.

@rcollyer Yep, I just figured out that there is actually a Span variant and decided to experiment. I think use of Span as such a list specification is interesting at least in cases where time complexity of the function is essentially independent of the "length" of the "list"...

Is it only my selective perception or do we have a lot of new, allegedly female users recently?

@kirma yeah, I had no idea it was valid until I looked at the docs, too. I agree, I think it's a good use of span.

@kirma nice, I didn't know it works with Span!

@kuba @Szabolcs I have now posted my stand on "bug-reporting" on Meta.

@halirutan Yes, one might well note that. It may be the name (legacy: Mathematica) as there are almost as many women studying Mathematics in Germany as there are men. Quite opposed to Computer Science... WRI: Don't change the name! :)

Hi everyone,
I have a system of PDE's.

PDE1 = D[u0[x, t], t] - Exp[-x] == 0

PDE2 = D[u1[x, t], t] + D[u0[x, t], x] -
   2*D[u0[x, t], {x, 2}, {t, 1}] + Exp[-x] == 0

PDE3 = D[u2[x, t], t] + D[u1[x, t], x] -
   2*D[u1[x, t], {x, 2}, {t, 1}] == 0

PDE4 = D[u3[x, t], t] + D[u2[x, t], x] -
   2*D[u2[x, t], {x, 2}, {t, 1}] == 0

PDE5 = D[u4[x, t], t] + D[u3[x, t], x] -
   2*D[u3[x, t], {x, 2}, {t, 1}] == 0

PDE6 = D[u5[x, t], t] + D[u4[x, t], x] -
   2*D[u4[x, t], {x, 2}, {t, 1}] == 0

ics = {u5[x, 0] == 0, u4[x, 0] == 0, u3[x, 0] == 0, u2[x, 0] == 0,

@rcollyer RandomChoice works too, but it's not documented...

Solved my Workbench build problem -- I had changed my paclet version to 0.9 from 1.0.0 in PacletInfo.m, which turns out to be a bad idea.

@gwr Or it is a result of the whole feminism going on lately. More girls to nerdy subjects!

Further investigation: removing the old docbuild.xml let me "downgrade" my paclet to 0.9.0

So - all is good

@kirma interesting. I've never looked through that code, so I think it would be interesting to find out how they manage the non-replacement issue with RandomSample. Bloom filters would be one option.

@ChrisK That shouldn’t have been a problem. But ah well it’s a finicky system. Does it make 11.3 style docs yet?

@rcollyer Probably depends a lot on what kind of time complexity is considered "good enough."

@halirutan I do not believe so, as there is no increasing trend for women to become "NERDS" - as I said, they probably are lured here simply by the misleading name. "Wolfram|One" will prevent this from happening.

@gwr OK, I forgot the <irony> tags around my message :)

@halirutan It is only your selective perception. You (and a few others) might consider getting more into the 21st century. And maybe Wolfram's marketing should start targeting that relatively untapped market (women not "girls"). Maybe that's why MATLAB and Maple seem more popular?

@JimB What happens in this 21st century that we are not aware of?

In short I think it's being aware that words chosen might be dissuading some very good minds from participating in this forum. Do I think that you guys show any animosity towards folks that ask questions here that depends on their gender? Absolutely not! You even treat us statisticians well (which shows a great deal of patience). My comment is simply to suggest thinking about word choice when gender is an appropriate topic in this forum.

Matrix math the easy way:

@b3m2a1 I can't tell the difference between 11.2 and 11.3-style docs.

The ones I make are pretty close to those, with a few minor differences -- most notably, instead of "Details and Options" I have "Details" and there is no thumbnail image encouraging opening that section

instead of orange cell openers that look like > and V, I have gray triangles

@halirutan Flashback‌​!

Never say anything... of anything. Solves the problem on fear of insulting somebody. Also makes sensible communication next to impossible with everybody.

Indeed, some would actually prefer womyn or wimmin instead of women, but somehow we have decided that it is ok to disregard them.

@C.E. I'm just glad Silvia took my comment for what it was: a high appreciation of her skills.

@kirma maybe I should just search SO for it.

@JimB I appreciate your clarification. I won't comment on it though. The cultural difference between Europe and the U.S. are just too large.

@C.E. Did you scroll up to the message I was replying? Silvia posted a photo of her and said "I had lunch with Daniel (Lichtblau)"

@halirutan I'm sorry for linking to such an old conversation, I apologize. I shouldn't have done it. I know the conversation and I reread it in its entirety though. I was only pointing out that JimB is not the only native speaker who doesn't like to use the word "girl" for women.

@C.E. Yeah, that's plain language difference. For me, boy/girl (in no specific order) identifies the more the gender than saying "young female".

I understand. "Girl" and "woman" cannot be used interchangeably in Swedish, so I am helped by that.

@C.E. In German neither. Bahh, I guess it's complicated. It just sounds friendlier.

I've been several times to the Beijing region and at some point, there was even some Mma-related occasion there when I was around... meeting Silvia might have been interesting, but usually my visits end up lacking slack and my better half might have a say on such things, too! ;)

@ChrisK Oh that's lazy on WRI's part. I actually meant the web docs though (just forgot to mention that). As of this past summer the docs it built by default were for 11.0.

That is, driving around the capital area can take a significant amount of time for relatively little shared interest. :I

@b3m2a1 I left comments under your pull request. Have you seen them?

@halirutan Yeah getting to them now. Didn't see them until late yesterday.

@b3m2a1 You know how you have to proceed with a PR?

@halirutan I think so. This is only my second, but in general I figure I just address whatever comments, commit necessary changes and then you merge or not as necessary.

@b3m2a1 Right, just change your local branch, make a new commit and push it. The rest will happen automatically.

I don't insist on removing the cache and the unnecessary window settings, but while we're at it, we can fix this along the way.

Yeah it's easy enough

@b3m2a1 I haven't bothered making HTML docs yet, still need to finish writing all the required Symbol pages and a few tutorials.

Is anyone here using ExternalEvaluate seriously? 11.3 has some improvements. What's the best way to pass arguments to a Python function? Use ExportString and build pieces of code?

@Szabolcs I wrote a Mathematica to python converter here that is reasonably robust. It uses a type-registration system to determine how to export arguments. Doesn't handle big data like images particularly well, but that's why I used a system for encoding them as JSON to pass between a python subprocess and Mathematica.

Thanks! I'll take a look.

For small stuff it works fine though:

ToSymbolicPython[
  Function[{x}, x + Range[100]] /@ Range[10]
  ] // ToPython

"list(map(( lambda x: ( x+range(100) ) ), range(10)))"

If it looks like it could be extended to suit your needs I'm happy to put in a few hours extending it.

Be warned that it makes pretty ugly python expressions in general:

ToSymbolicPython[
   # + Range[100] & /@ Range[10]
  ] // ToPython

"list(map(( lambda _lambda_var_1: ( _lambda_var_1+range(100) ) ), \
range(10)))"

@b3m2a1 I was thinking about the simplest way to call some networkx functions.

Here's a kinda clumsy way to do that:

ToSymbolicPython[
  Import["networkx" -> "nx"];
  G = networkx.Graph[];
  G.add$edge[1, 2];
  G.add$edge[2, 3, weight -> 0.9]
  ] // ToPython

"import networkx as nx
G = networkx.Graph()
G.add_edge(1, 2)
G.add_edge(2, 3, weight=0.9)
"

Let me remember how you register specific type conversions and then I'm sure I can find a way to auto-serialize a Graph using what I have already...

So it seems I cooked all pre-conversions into a symbol called $PyLangTranslations. So you can register a definition that will turn a valid Graph object into a symbolic python construct like:

AppendTo[
  $PyLangTranslations,
  g_Graph?GraphQ :>
   RuleCondition[PyTuple@*List @@@ EdgeList[g], True]
  ];

And then this works

With[{g = RandomGraph[{10, 10}]},
 ToSymbolicPython[
   Import["networkx" -> "nx"];
   G = nx.Graph[g]
   ] // ToPython
 ]

"import networkx as nx
G = nx.Graph([ ( 1, 2 ), ( 1, 10 ), ( 2, 6 ), ( 2, 8 ), ( 3, 4 ), ( 3, 5 ), ( 3, 10 ), ( 4, 8 ), ( 4, 9 ), ( 5, 6 ) ])
"

You have to insert the graph because ToSymbolicPython is HoldAll.

I probably should have used a slightly more robust format than a list of replacements as to deregister the rule you need to use the clumsy:

$PyLangTranslations =
  DeleteCases[$PyLangTranslations, Verbatim[g_Graph?GraphQ] :> _];

You could also build out a replacement that pre-evaluates any valid Graph constructors like:

Ah wait that one's a bit nastier owing to order of ReplaceRepeated operations. I'm gonna need to specify some pre-conversion evaluations I think.

Okay in the version I just pushed you can add rules to $PyLangPreEvaluate to specify pre-conversions, like:

PrependTo[
  $PyLangPreEvaluate,
  g : _GridGraph | _LineGraph | _PathGraph | _StarGraph | _TreeGraph \
| _CycleGraph | _IndexGraph | _TuranGraph | _WheelGraph | \
_CayleyGraph | _HararyGraph | _RandomGraph | _SimpleGraph | \
_CompleteGraph | _DeBruijnGraph | _PetersenGraph | _ButterflyGraph | \
_CanonicalGraph | _CirculantGraph | _HypercubeGraph | \
_KnightTourGraph | _UndirectedGraph | _TransitiveClosureGraph | \
_TransitiveReductionGraph :> With[{gg = Evaluate[g]}, gg /; GraphQ[gg]]
  ];

And then this will work:

ToSymbolicPython[
  Import["networkx" -> "nx"];
  G = nx.Graph[WheelGraph[10]]
  ] // ToPython