Dear Bertrand, Thanks for your objection, and of course your example is clear and sound. Please, let me please clarify. IMHO, the inheritance relation in Eiffel covers both: an inheritance of code and contracts from the parent to the child, and a subtype relationship from the child to the

Hi Colin, What about using `start_pos = lower or valid_index (start_pos)` instead? It's a bit stricter than what you suggest but has more meaning to me: the caller tries to get a subarray, and if the current array is empty, the subarray should be requested with the same lower index rather than

Revised Iteration Benchmarks I have revised my benchmarks article on eiffel.org when I realised I made a mistake and was only doing a partial garbage collect after each benchmark iteration run. Now with a full collect do_all is marginally (but consistently by 1%) beating across. Also the results

Hi Larry, I have revisited my benchmarks again and written this article on eiffel.org which compare 5 different iteration styles and examines the difference caching the array count as a local variable makes. We seem to be getting quite different results especially on do_all which ties with

Runtime Conditions EiffelStudio version: 18.1 (18.01.10.1424 GPL Edition - win64) C Compiler: Microsoft Exe type: finalized exe Code inlining: (inlining size = empty) OS: Windows 10 Larry Rix Moonshot Software Rocket science for everyone! Savannah, GA 770-295-9729 On Fri, May 18, 2018 at

It would be interesting to see what effect setting inline size to 2 would have on your results. I just realized I was calling MEMORY.collect instead of MEMORY.full_collect after each benchmark. Will need to recompile and run again. On Friday, 18 May 2018 17:37:16 UTC+1, Larry Rix wrote: > >

Hello Irix, BTW: across would be in Ruby: [ 1, 3, 5 ].each { |i| puts i } Perhaps Ruby is something for the Rosetta pages? Gerrit Woland's Cat schrieb am Fr., 18. Mai 2018 um 09:11 Uhr: > > That list may have added to it lambdas. In Eiffel, function currying > works, so

Auf: eiffel...@googlegroups.com am Friday, May 18, 2018 11:22 sprach mein guter Freund Bernd Schoeller also: > Sorry to be a little harsh, but "replicating inheritance" is IMHO a broken concept that should not be used. >.[…] >PS: A classic example for replicating

Hi Louis, Sorry to be a little harsh, but "replicating inheritance" is IMHO a broken concept that should not be used. Replicating inheritance always happens when you have a single parent concept multiple times in the same class. In Eiffel, it is very simple to detect, as it forces you to use the

That list may have added to it lambdas. In Eiffel, function currying works, so that an agent with partly filled arguments generates a new agent with a new signature; this doesn't work in Java 8, you just have fixed lambdas (which are neverthelss pretty useful, compared to previous versions of

I think it might be possible, but (of course) we want people flowing from Other-language-to-Eiffel and not so much the other way around (e.g. we're not here to teach Java to Eiffel programmers, eh?) :-) In the reverse view: Eiffel Java DbC No, but there are asserts MI No user infix ops No

Hi Larry, Thanks for the information. It is quite a nice format as it goes - for ‘notes’. But for more serious, published document, monospaced typefaces are a big step back to the early 1980s. I agree content is the most important, but that is not a reason to ignore the formatting (or an

I think it might be possible, but (of course) we want people flowing from Other-language-to-Eiffel and not so much the other way around (e.g. we're not here to teach Java to Eiffel programmers, eh?) :-) Larry Rix Moonshot Software Rocket science for everyone! Savannah, GA 770-295-9729 On Thu,

Anthony, I agree that a Rosetta-code with strong examples that "work-the-first-time-every-time" would accelerate adoption of Eiffel into mainstream projects. I will have to look into the fastest and best roads to getting Eiffel-to-C#.NET hooked up. Larry Rix Moonshot Software Rocket science for

Larry, this is the standard form. Although I tend to use xx_csr as the iterator name, as in across thing_list as thing_csr do ... thing_csr.item end - thomas

I can only add my perspective as someone who has used C# and Java for many years, and has dabbled in Eiffel off and on over the years with a desire to use it as a primary tool. I can figure out the majority of the language and constructs on my own, but translating common idioms/programming

Finian: Great idea! Is it possible to have two-way tables so that when programmers see Eiffel constructs they might reference the Java analogue to help their comprehension with the Eiffel? Jonathan On May 17, 2018, at 8:38 AM, Finnian Reilly wrote: Quickly Orientating

All this is extremely interesting. The contributors are expert Eiffel users. I would like to make sure that we keep the needs of novices in mind. If an iteration over 1 |..| 10 is not optimal, as Alexander Kogtenkov pointed out, this is important for advanced users (for whom, presumably, the

I was noting some wild differences in my benchmark code runs. I noted that benchmark times were wild in Workbench mode and changed most greatly if I did a freeze. That led to me just finalize the code to an EXE and try again. Once the code was in finalized, the wild swings continued. At this

Furthermore, you should compare loops that do something (the same thing, but something). Loops that do nothing can be optimized away by the underlying C compiler. -- Eric Bezault mailto:er...@gobosoft.com http://www.gobosoft.com

They are expensive if the GC collecting everything that you created before is triggered at that time. You should run your benchmarks separately, not one after the other in the same execution. -- Eric Bezault mailto:er...@gobosoft.com http://www.gobosoft.com

Ah, good to know! Then perhaps this is not just OS related but simply the GC getting involved as well, yes? On Thursday, May 17, 2018 at 11:01:13 AM UTC-4, Eric Bezault wrote: > > On 5/17/2018 16:54, Larry Rix wrote: > > * Going across l_container is just as fast on 1_000 INTEGERs as the > >

Totally agree that the point of this tasking is to train at a high level and leave the granular and more nuanced matters to later material. Perhaps an entire section of docs that deals with performance issues and benchmarking both within Eiffel and between Eiffel and other language results for

Looks like you are measuring performance using workbench mode. That way lies madness. On Thu, 17 May 2018 at 15:54 Larry Rix wrote: > I lowered the l_upper_bound from 1_000_000 to just 1_000. The results are > interesting. > > The across loop (in certain forms) is

The Eiffel compiler ought often to be able to optimise across over an array by compiling to an integer index loop instead. On Thu, 17 May 2018 at 15:33 Larry Rix wrote: > Hi again, Finnian, > > I am going to modify my statements about across being slower. As I was >