
Re: Work on Basic Mathematica Stephen!
Posted:
May 16, 2013 3:28 AM


Well, lots of things can be proposed and everyone has different ideas.
I think that what's really needed for WRI is not to get too distracted by various proposals, but *focus* on doing a few things well, and especially focus on fixing bugs.
Primarily silent wrong result bugs.
The two nastiest examples (that affected me) tat I can recall are:
Tally[] with a custom test was broken in version 6, and gave wrong results silently. It wasn't fixed in minor releases, we had to wait 2 years for 7.
IsomorphicGraphQ[] was broken in 8 and gave wrong results silently. It also crashed sometimes. The crash was fixed in a minor version, but the silent wrong result (which I consider a much worse bug) was not fixed until v9 two years later.
These silent wrong results are very serious problems because they are completely unexpected and should be fixed as soon as they are discovered.
They are not at all in the same class as an incorrect result for a definite integral with a singularity, which is expectedno CAS will be perfect in this. Such a bug in either Tally or IsomorphicGraphQ is not expected
On 15 May 2013 05:04, Peter Klamser <klamser@googlemail.com> wrote:
> This is an interesting discussion. But if it can be useful, we have to > make short proposals. Nobody has the time to read long texts. > > A) First proposal: Identify useless or false constructions in Mathematica > > aa) Eliminating Chop[] for complex numbers. Complex numbers are > oderless and therefore nobody call estimate, weather the distance of 1 > + 10^google i to the real numbers is small or big. Chop[] is the > result of Mathematica design, that it presents often complex results, where > real values are the simpler result and can be reached by > ComplexExpand[]. > The simplest solution is always the best solution. > > B) Mathematica should be integrated seamless in Tex without using copy and > paste. The should be the cell type Tex, and the printout and save > modus Tex, where all other cells are suppressed. > > Peter > > > > 2013/5/15 Szabolcs Horv=C3=A1t <szhorvat@gmail.com>: > > > > While some individual points are debatable, I also completely agree > > with David's main message. Mathematica needs more focus, particularly > > some work on stability and robustness. Even if it would come at a cost > > of new features (there are many features I'd like to see, but you can't > > have everything). > > > > Re: "if I click in an existing Input cell and do a line return the > > Messages window opens with a contact WRI if this happens message" > > > > This bug (on OS X) is really annoying when typing code in another > > language in a Mathematica string. If will sometimes completely ruin > > the typed text if I press return several times while inside a string. > > It's a bit ironic that this bug appeared in the same release which > > brought us RLink (which I'm using these days, i.e. I'm typing a lot of > > R code inside Mathematica strings). > > > > On 20130512 07:28:34 +0000, djmpark said: > >> (I've renamed this and started a new thread because my reply is not > exactly > >> to the question.) > >> > >> Oh, what a wonderful Wolfram blog! Earlier Stephen hinted at > Mathematica as > >> an iPhone app. Now it's data mining Facebook data (Gee I wonder if > >> Zuckerberg has thought of that? He might be able to develop a great > business > >> model.) Can Twitter be far behind? There are many significant > mathematical > >> equations that will fit into 64 characters  or whatever the limit is. > >> Ramanujan would probably have done well on Twitter. And women are more > >> interested in personal relationships and men are more interested in > sports? > >> Who would have thought? The average person on Facebook has 342 friends= ! > Well > >> there are friends and there are friends. Montaigne wrote that his > friendship > >> with Etienne de La Bo=E9tie was such that "So many coincidences are > needed to > >> build [it up] that it is a lot if fortune can do it once in three > >> centuries." One might say, ephemera in ephemera out. > >> > >> For the dwindling few of us who still have desktop computers and large > >> screens, or maybe two large screens, who are interested in learning or > doing > >> some extended mathematics, and the even fewer who would like to write > >> literate Mathematica notebooks as technical documents, I wonder if > Stephen > >> could find some time to attend to basic Mathematica, fixing its > problems and > >> fulfilling its vision? > >> > >> Mathematica lacks stability. Things that worked fine in one version > don't > >> work in the next. Especially troubling to me is the basic user > interface. > >> This got much worse in Version 9 with outright bugs that are in your > face > >> all the time. For example: often if I click in an existing Input cell > and do > >> a line return the Messages window opens with a contact WRI if this > happens > >> message. Or if one clicks after a word in a Text like cell and uses > Ctrl+K > >> for spell checking the message window again opens with a similar > message. I > >> like to use spell check a lot so this is especially annoying to me. I > don't > >> see why WRI couldn't have fixed these problems by now. (Or introduced = an > >> actually useful feature to spell check a selection such as a Text cell > or a > >> Section.) > >> > >> The Version 9 command completion feature, which used to be great, no > longer > >> works well. I have turned off the auto completion feature but what is > left > >> still does not work as well as the Version 8 behavior. For example, i= f > one > >> types: > >> > >> Carm > >> > >> and then uses Ctrl+K to complete the command, there is only one choice= , > >> CarmichaelLambda. In Mathematica 8 the symbol would be automatically > >> completed and the cursor would be left at the end of the word. But now > >> it brings up a menu, > >> even though there is only one choice, and one must click the menu. But > you > >> are not finished yet! No there is another menu (somewhat displaced so > you > >> may miss it), which is the equivalent of Ctrl+Shift+K and which we cou= ld > >> have done if we wanted, so one must bypass that by clicking at the en= d > of > >> CarmichaelLambda to get back to the normal typing entry. That's two > extra > >> clicks added. You might say that's not much, but when it's at the basi= c > >> entry point for material in a notebook it is a lot. > >> > >> One has to wonder how many parsers there are in Mathematica for kernel= , > >> front end, packages, workbench and how their behavior shifts around > between > >> versions? Can one copy and paste an expression without its underlying > >> representation changing? I suspect this may be a nagging underlying > source > >> of instability. > >> > >> I realize and appreciate that WRI continues to add new capabilities to > >> Mathematica and this inevitably results in learning and stability > problems. > >> There could be better design efforts on these things and more > professional > >> testing so the designs would stick and work well. Progress might be > slower > >> but it would be surer. > >> > >> Doing mathematics is not social media. It's not done that well on an > iPad. > >> And iPhone, iPad technology is not necessarily appropriate for > Mathematica. > >> People do not want to scroll two 25" screens with their hands. Just as= k > >> Microsoft. > >> > >> Mathematics and Mathematica are intrinsically difficult enough as it i= s. > >> That makes it all the more important that WRI present users with a > stable, > >> robust, easy to use basic interface. WRI had not done the best job the= y > >> could at this and not what one would expect for a relatively expensive > >> product. > >> > >> > >> David Park > >> djmpark@comcast.net > >> http://home.comcast.net/~djmpark/index.html > >> > >> > >> From: r.b.nachbar@gmail.com [mailto:r.b.nachbar@gmail.com] > >> > >> > >> I'm interested in making word cloud graphics, as demonstrated in the > recent > >> Wolfram Blog post > >> http://blog.wolfram.com/2013/04/24/datascienceofthefacebookworld/= (near > >> the end of the post). Does anyone have a function that will do this? > >> > >> Thanks, > >> > >> Bob > > > > > > >

