Talk:Arbitrary-precision integers (included): Difference between revisions

← Older edit

Talk:Arbitrary-precision integers (included) (view source)

Revision as of 18:22, 2 June 2021

13,773 bytes added , 2 years ago

m

added a comment.

Anonymous user

rosettacode>Gerard Schildberger

Revision as of 01:40, 9 May 2011 (view source) rosettacode>Kernigh (→‎Missing output: new section) ← Older edit		Latest revision as of 18:22, 2 June 2021 (view source) rosettacode>Gerard Schildberger m (added a comment.)
(34 intermediate revisions by 8 users not shown)
Line 1: ==Use of external libraries== I ''strongly'' recommend allowing the use of external libraries in the case where the language doesn't have inherent bignum support. Otherwise, you would need a more generalized task, and I don't see a point to the example code duplication that would require. --[[User:Short Circuit\|Michael Mol]] 08:33, 13 February 2010 (UTC) Line 20 ⟶ 21: ::: Yeah, there isn't an official library. The forum and the wiki are it. Laszlo is the goto person for math related things in autohotkey. I have added a link to it on the [http://www.autohotkey.com/wiki/index.php?title=Script_Listing#Math_.26_Calculators wiki] as well. ::I would suggest then that if Laszlo's library is available at a standard place, then the AHK example could just refer to the library and show its use in solving the problem rather than quote the library on RC, as, for AHK at least, that would then make Laszlo's bignum library a "standard library"? --[[User:Paddy3118\|Paddy3118]] 06:06, 6 June 2010 (UTC) Late answer... I agree that external libraries should be allowed, and it make little sense to require ''a single, overwhelming, library of varied modules that is endorsed by its home site''. In some cases, the language does not have a single home page, simply because it's an international standard. In Fortran, there are at least 3 well known multiprecision libraries (by [https://maths-people.anu.edu.au/~brent/pub/pub043.html Brent], [http://crd-legacy.lbl.gov/~dhbailey/mpdist/ Bailey] and [http://myweb.lmu.edu/dmsmith/fmlib.html Smith]). Should we dismiss all of them because it is required to have only one ''overwhelming'' library? As Michael Mol commented above several years ago, that means we would need another more general task, and duplicate code. The task looks currently too rigid to me. For the record, I have added a solution using Smith's FM, and I may add solutions with the other two later. [[User:Arbautjc\|Arbautjc]] ([[User talk:Arbautjc\|talk]]) 23:58, 11 November 2016 (UTC) :Ah well. It seems to me that the point of the task is to demonstrate the ''in-built'' facilities for arbitrary-precision integers, with however allowance for the existence of some library extension so routinely available at every installation that everyone would already be using it for such tasks. If the required number size had not been so large, the IBM1620 would have managed, but perhaps modernists with ibmpc-monocuture aren't interested. Evidently, any language can be extended to present such facilities, but the specification explicitly rejects developing such a library. If in Fortran's case there are three such libraries, and (as it appears from the above) different verbiage is required to employ each for the same calculation, we would find ourselves comparing such libraries rather than the in-built features of different languages. [[User:Dinosaur\|Dinosaur]] ([[User talk:Dinosaur\|talk]]) 10:29, 13 November 2016 (UTC) ::True. But then remove also the C and Ada entries, and probably others as well. Oh, wait, the point of this comment was to show that it would be necessary then to add a task to allow external libraries, since many languages don't have builtin multiprecision. It was already stated by Michael Mol in 2010 above. Now, what's most useful? A long talk about the IBM 1620, a computer from 1959 that has nothing to do with modern Fortran, or a way to do multiprecision computations in Fortran today? I am interested in history, but I am definitely interested first in practical way to do things right now: programming is not primarily about history. But, if really asking to slightly change the task (so that some entries become conformant, and not only Fortran), is too much, then I'll copy-cut to create a new task to allow this. I don't think it's particularly wise though. [[User:Arbautjc\|Arbautjc]] ([[User talk:Arbautjc\|talk]]) 11:41, 13 November 2016 (UTC) :::I note the remarks in the Algol68 example about strictly speaking, and so therefore, those entries that involve external features or explicit routines for performing multi-precision should indeed be ejected as not conforming to the task's statement, strictly. Otherwise, we face mere variations of something like <include library>...<eval(5^4^3^2)> with specious differences not elucidating the language facilities. Entries for the likes of Alore and bc amount to that but without the "include" because the necessary facilities are built-in to the language. Only some libraries amount to being a part of the language, as with C's stdio.h. I'm interested in the historical changes in a language as features are introduced or removed and generality is attained or implemented with limits - so no INTEGER2000 for example, even though in principle this could be allowed, somewhat as in Algol68 - and was amused by the thought that the IBM1620 could perhaps have succeeded, and the IBM1403 also? These represent a possible development path in this matter that was not taken and anyone interested in language comparison would be interested in such decisions. As for what's more useful, someone seeking to solve such a problem "right now" via perusal of RC to select a language for immediate use would want to know which languages had built-in facilities or not. And reading the task description, would expect to see just that, and if for example bc was available, success would be moments away. But with libraries required there would be a delay, unless the library was overwhelmingly available so as to seem a part of the language and even more delay for supplied routines. Thus I am speaking against the inclusion of libraries or specially-written routines, as per the task specification for this specific task. :::As for organising another task with or without this or that, I leave that to others as already, many flowers bloom and I wander without guide. There does not seem to be a facility for grouping related tasks, or multiple parts of a larger task, or even cross-referencing tasks. Just a flat list of tasks ordered alphabetically and with not always helpful choices of the first word. A tree-structure? But the A.B.etc, B.X.etc ordering won't work when it is not clear which should be the first-level keywords. This is a matter for the design of RC's presentation, which I know little about, and I see a lot of scope for disagreement. [[User:Dinosaur\|Dinosaur]] ([[User talk:Dinosaur\|talk]]) 07:09, 14 November 2016 (UTC) ::::Just for fun: even languages with builtin multiprecision may have additional libraries, for good reasons. For instance, Python has big integers, but the gmpy2 library is linked to GMP, which is way faster than the standard Python integers (plus it also adds multiprecision floats, which are not in the base Python distribution). It makes very little sense to only allow what is included in the bare language: who programs in C++ without BOOST? Who programs in Fortran without LAPACK? Who programs in Lisp without ASDF or even QuickLisp and a bunch of libraries? Who programs, in any language, a GUI without a GUI library? (rarely included, and even when it is, it's never the only one) It's a basic fact of programming that one should not reinvent the wheol but reuse, hence it would be perfectly correct to link to an external library. That's how it ''should'' be done in real life. But, bitter truth, this is not the present task. I'll write another one, maybe. This is really silly. Feel free to quibble about what Fortran could have been sixty years ago, it's almost as useless as the rest of the task! [[User:Arbautjc\|Arbautjc]] ([[User talk:Arbautjc\|talk]]) 18:07, 14 November 2016 (UTC) Interestingly, because the fixed const value is needed, the perfect solution may be just one liner hardcoded ''print("62060698786608744707 ... 92256259918212890625")''. And the "bad" solution is any implementation of Muḥammad ibn Mūsā al-Khwārizmī algorithms. '''The task has is not well-defined.''' Different programming languages are different. Some are monolithic (such as Pascal). Others are modular by design. Therefore, the task should be modified. It should be allowed to use external libraries (from generally available repositories such as PyPI, CPAN etc.) - and '''when there is no arbitrary arythmetic''' in such standard libraries then showing how the goal can be achieved using standard language mechanisms. Standard mechanisms '''should be understood as ''write functions, procedures, classes, etc., to perform what is necessary.''''' Anyone seeing the source code will be able to '''judge for themselves''' how convenient a given language is for this type of calculation. --[[User:Anonymous31415927\|Anonymous31415927]] ([[User talk:Anonymous31415927\|talk]]) 11:49, 12 September 2020 (UTC) :The task notes address this. Why do you think them inadequate? --[[User:Paddy3118\|Paddy3118]] ([[User talk:Paddy3118\|talk]]) 19:42, 12 September 2020 (UTC) First of all, the task should be redefined to '''enforce computations during the run time''', because 5(4(32)) can be precomputed or computed in a compile time (esp. in Haskell-like). Secondly, there is no need to say ''"do not submit an implementation of arbitrary precision arithmetic".'' De facto, to solve this task some kind "arbitrary precision arithmetic" (20 digits at least) must be used- either as an internal feature or as an external module/library. Finally, see "pure C solution" - it is '''ugly, but it SOLVE''' the problem: it is criticized as "this example is incorrect". It is not justified in any way, because the Rosseta Code is not a code golf competition and the length of the source code is not revelant. --[[User:Anonymous31415927\|Anonymous31415927]] ([[User talk:Anonymous31415927\|talk]]) 12:41, 17 September 2020 (UTC) == Criteria for Non-Draft? == Line 26 ⟶ 54: :I'm pretty sure there are no criteria. It's just a consensus. This one seems to still have a little discussion going on about how we want it to work so maybe we should wait to make it a full task. --[[User:Mwn3d\|Mwn3d]] 16:45, 13 February 2010 (UTC) ::Whoops. I've changed it already due to the number of examples. Please put it back in draft if you think it needs it though. --[[User:Paddy3118\|Paddy3118]] 16:58, 13 February 2010 (UTC) ==“Precision”== Line 33 ⟶ 62: ::''"In computer science, arbitrary-precision arithmetic is a technique whereby calculations are performed on numbers whose digits of precision are limited only by the available memory of the host system. This contrasts with the faster fixed-precision arithmetic found in most ALU hardware, which typically offers between 6 and 16 decimal digits. It is also called bignum arithmetic, and sometimes even "infinite-precision arithmetic" (which is a misnomer, since the number of digits is both finite and bounded in practice)."'' : --[[User:Paddy3118\|Paddy3118]] 16:54, 13 February 2010 (UTC) ==Ursala took a day?== On my four year old laptop, the Python version took around a minute to run. Maybe the Ursala version should not have used BCD if another representation is available? --[[User:Paddy3118\|Paddy3118]] 19:08, 8 April 2010 (UTC) == Time constriction? == ---- Copied from [[User talk:Short Circuit]]: Line 49 ⟶ 82: == Does the Clojure entry count the digits? == --[[User:Paddy3118\|Paddy3118]] 06:13, 11 June 2010 (UTC) ==Move timings to the talk pages== I'm not sure that timings should be included in answers except maybe in general terms, as [http://shootout.alioth.debian.org/ The Computer Language Benchmarks Game] does it in a more rigourous way and yet still points out [http://shootout.alioth.debian.org/dont-jump-to-conclusions.php issues]. --[[User:Paddy3118\|Paddy3118]] 00:34, 25 February 2011 (UTC) Line 57 ⟶ 93: ::Thanks Kernigh. I know I prefer it this way rather than having to out-do TCLBG. What do others think? --[[User:Paddy3118\|Paddy3118]] 04:05, 27 February 2011 (UTC) === Timings with AMD Phenom II X2 560 === # C with GMP: 0m0.04s # Ruby 1.9.2: 0m0.52s Line 67 ⟶ 105: --[[User:Kernigh\|Kernigh]] 03:26, 27 February 2011 (UTC) == Missing output == Line 75 ⟶ 114: There should not be tag like <nowiki>{{incorrect\|Golfscript\|Output not shown.}}</nowiki> unless the code is missing a print command. --[[User:Kernigh\|Kernigh]] 01:40, 9 May 2011 (UTC) :Hi Kernigh, the task is written that the output is to be shown. People mis-interpret the task and write ''some'' entries without output. That does not mean that all output should not be shown from then on. I try and keep entries on track by reminding people of the need for output to be shown. :I do note that it is very rare for people to misunderstand a task and insist on doing ''more'' rather than less. By arguments such as yours a task could get steadily worse over time as people leave out one or more aspects over time and others look to implement the bare minimum. :In this case, could you not remove the need for output to be shown until someone actually shows the output from the programs? Thanks. --[[User:Paddy3118\|Paddy3118]] 05:26, 9 May 2011 (UTC) :Strange? You added output for Scala, but just removed the templates for Factor and GolfScript. --[[User:Paddy3118\|Paddy3118]] 05:30, 9 May 2011 (UTC) We are confusing two different ideas: # Show the output to the user running the program. # Post the output to this wiki page. Users can post the output to this wiki page, but that is no task requirement. A language that shows the output, but does not post to this wiki page, is not wrong. For example, the [[Loops/Foreach]] task says to "print each element", but few languages post the output to that wiki page. --[[User:Kernigh\|Kernigh]] 22:17, 15 May 2011 (UTC) : Seems fair to me. To be exact, so long as the right output is produced, it's not required here. Especially if there is model output that can be pointed at. –[[User:Dkf\|Donal Fellows]] 00:03, 16 May 2011 (UTC) : Unfotunately, that interpretation is wrong. The only reason for: :# Duplication/emphasis of print-like keywords e.g. "print and show" or "generate and show" in this case it is "find and show". :# Attaching an <nowiki>{{output?\|language\|reason>}}</nowiki> box on examples that don't. :# Having the task author repeatedly state in the task talk page that output should appear. :# Having edits summaries for revisions mentioning the lack of output. :# The majority of other examples having output shown. : ... Is to guide fellow contributors to the conclusion that the giving of output is a necessary part of the task description. : Note also that in case of personal doubt, you are less likely to need to revise your submission if you include the output. --[[User:Paddy3118\|Paddy3118]] 05:15, 16 May 2011 (UTC) ==Undertested cosmetic edits at 05:58, 24 July 2016 hid formula from most browsers== Cosmetic edits made to the task page at 05:58, 24 July 2016 , including the injection of spaces around expressions in <math> tags, left the main expression completely invisible to all browsers which display the graphic file version of formulae rather than processing the MathML (this is, in fact, the majority of browsers). The MediaWiki processor does not currently expect such spaces, and generates syntactically ill-formed HTML if they are introduced. Other aspects of these cosmetic edits may further compound the problem. [[User:Hout\|Hout]] ([[User talk:Hout\|talk]]) 08:48, 21 September 2016 (UTC) : Repaired [[User:Hout\|Hout]] ([[User talk:Hout\|talk]]) 00:29, 3 November 2016 (UTC) ==failure of ooRexx using Classic REXX example== (Moved here from the epilogue (after the '''output''') of the Classic REXX example.)   -- [[User:Gerard Schildberger\|Gerard Schildberger]] ([[User talk:Gerard Schildberger\|talk]]) 18:31, 10 September 2017 (UTC) <pre> ooRexx gives me 5 -* n=5 (4 (3** 2)) /calc. multiple exponentiations. / Error 26 running Z:\huge.rex line 5: Invalid whole number Error 26.8: Operand to the right of the power operator (**) must be a whole number; found "2.6214E+5" Other Rexx-es don't?? </pre> No answer yet. numeric digits 6 at the start fixes the problem for ooRexx --[[User:Walterpachl\|Walterpachl]] ([[User talk:Walterpachl\|talk]]) 18:47, 5 November 2013 (UTC) : (Note:   the above comments should've been placed under the ooRexx language section, or here, on the discussion page. : Also note that ooRexx won't execute a lot of Classic REXX examples   (as evident here on Rosetta Code)   as they are different languages with different syntax (dialects?) and statements,   despite having '''REXX''' in their names.   Although ooRexx claims to run any Classic REXX, all evidence is to the contrary. : Yes, Regina and R4 are both very lenient (liberal?) in their usage of exponents.   But increasing the numeric digits just bypasses the problem, it doesn't fix it in that it is apparently one more difference between ooRexx and Classic REXXes. : Also, the (internal) subroutines of these REXXes may use a   '''numeric digits 9''',   thereby bypassing the use/limit   '''numeric digits 5'''   (as in the older/replaced 2<sup>nd</sup> REXX version).     -- [[User:Gerard Schildberger\|Gerard Schildberger]] ([[User talk:Gerard Schildberger\|talk]]) 18:31, 10 September 2017 (UTC)