Re: Fwd: FOSDEM - perl 6 critic
Moritz Lenz mor...@faui2k3.org writes: Am 22.02.2011 16:57, schrieb Gabor Szabo: He, as a sysadmin would like to do the small tasks in a relatively small language. He would like to make sure the modules/applications he will download and will have to support are in such a relatively small language. Whatever Perl 6 will turn out to be, it won't be a small language. And just to complement the many answers: There *are* small variants of Perl6: Perlito and NQP. Kind regards, Steffen -- Steffen Schwigon s...@renormalist.net Dresden Perl Mongers http://dresden-pm.org/
Fwd: FOSDEM - perl 6 critic
hi, At FOSDEM I met Arne Wichmann who is a long time sysadmin, Debian developer and Perl user. We had a short chat in which he expressed his concerns about the complexity, the size (memory footprint) and speed of Perl 6, Without even taking in account the current memory requirements and speed of Rakudo, I guess, even after lots of improvements we can expect Rakudo to be significantly slower than Perl 5.10 - at least for start-up time - and significantly more memory hungry. I know it will do a lot more so the comparison is not fair but that's not the point. ( For a better comparison that takes in account the features as well see http://www.modernperlbooks.com/mt/2010/07/an-accurate-comparison-of-perl-5-and-rakudo-star.html ) He, as a sysadmin would like to do the small tasks in a relatively small language. He would like to make sure the modules/applications he will download and will have to support are in such a relatively small language. I wrote him my opinion but I think it would be important to address these issues. (Of course if there already is a page somewhere answer these concerns I'd be happy to just get a link) Here is his e-mail. (forwarded with permission). regards Gabor -- Forwarded message -- From: Arne Wichmann a...@anhrefn.saar.de Date: Mon, Feb 21, 2011 at 4:00 PM Subject: FOSDEM - perl 6 critic To: ga...@perl-ecosystem.org Hi... You gave me your card when we were leaving FOSDEM, it took me some time to write a mail... The topic was: why I am very sceptic about perl6... First, my background: I am a perl hacker since '91 (or so), but mainly I am a sysadmin. That means, I do not write a lot of code, but I do a lot of debugging of other peoples code. From that background, what I have seen in perl6 does not look like a good idea to me: it is too complex. When I read other peoples code I have to be able to understand whatever subset of the perl language they choose to use - which means I have to be able to grasp any concept used in the language. And given the number and complexity of operators in perl 6 I do not feel that this is really doable. My other gripe is that perl5 nowadays already is too big - it takes too much memory and time for small tasks. But that is only secondary. cu AW -- [...] If you don't want to be restricted, don't agree to it. If you are coerced, comply as much as you must to protect yourself, just don't support it. Noone can free you but yourself. (crag, on Debian Planet) Arne Wichmann (a...@linux.de) signature.asc Description: PGP signature
Re: Fwd: FOSDEM - perl 6 critic
On Tue, 2011-22-02 at 17:57 +0200, Gabor Szabo wrote: For a better comparison that takes in account the features as well see http://www.modernperlbooks.com/mt/2010/07/an-accurate-comparison-of-perl-5-and-rakudo-star.html Thanks for posting this. Can I infer from this article that it is *rakudo*, which is slow, rather than parrot? I have the impression that I can compile perl6 down to parrot code and run that separately. If so, it will be interesting to benchmark both cases. -- --gh
Re: Fwd: FOSDEM - perl 6 critic
On Tue, Feb 22, 2011 at 6:14 PM, Guy Hulbert gwhulb...@eol.ca wrote: On Tue, 2011-22-02 at 17:57 +0200, Gabor Szabo wrote: For a better comparison that takes in account the features as well see http://www.modernperlbooks.com/mt/2010/07/an-accurate-comparison-of-perl-5-and-rakudo-star.html Just to clarify a bit to avoid misunderstandings. I am not criticizing the speed or memory footprint of Rakudo. I trust (blindly but still :-) the developers that they will achieve their goals. What I am concerned is that people who only get very limited information about Perl 6 will get (further) be turned off from Perl and Perl 6 due to whatever reason. The funny thing is that now that I actually followed the link I provided I can see I made a similar comment there. Gabor
Re: Fwd: FOSDEM - perl 6 critic
On Tue, Feb 22, 2011 at 11:14 AM, Guy Hulbert gwhulb...@eol.ca wrote: On Tue, 2011-22-02 at 17:57 +0200, Gabor Szabo wrote: For a better comparison that takes in account the features as well see http://www.modernperlbooks.com/mt/2010/07/an-accurate-comparison-of-perl-5-and-rakudo-star.html Thanks for posting this. Can I infer from this article that it is *rakudo*, which is slow, rather than parrot? I have the impression that I can compile perl6 down to parrot code and run that separately. If so, it will be interesting to benchmark both cases. -- --gh Parrot's speed memory footprint can certainly be improved - it's not all Rakudo. -- Will Coke Coleda
Re: Fwd: FOSDEM - perl 6 critic
On Tue, 2011-22-02 at 11:28 -0500, Will Coleda wrote: On Tue, Feb 22, 2011 at 11:14 AM, Guy Hulbert gwhulb...@eol.ca wrote: On Tue, 2011-22-02 at 17:57 +0200, Gabor Szabo wrote: For a better comparison that takes in account the features as well see http://www.modernperlbooks.com/mt/2010/07/an-accurate-comparison-of-perl-5-and-rakudo-star.html Thanks for posting this. Can I infer from this article that it is *rakudo*, which is slow, rather than parrot? I have the impression that I can compile perl6 down to parrot code and run that separately. If so, it will be interesting to benchmark both cases. -- --gh Parrot's speed memory footprint can certainly be improved - it's not all Rakudo. Ok. So we'll want to implement algorithms in both perl6 and parrot assembler then. -- --gh
Re: Fwd: FOSDEM - perl 6 critic
On Tue, 2011-22-02 at 17:35 +0100, Moritz Lenz wrote: Am 22.02.2011 17:14, schrieb Guy Hulbert: On Tue, 2011-22-02 at 17:57 +0200, Gabor Szabo wrote: For a better comparison that takes in account the features as well see http://www.modernperlbooks.com/mt/2010/07/an-accurate-comparison-of-perl-5-and-rakudo-star.html Thanks for posting this. Can I infer from this article that it is *rakudo*, which is slow, rather than parrot? There are three different sources for slowness: Thanks for the clarification. 1) parrot itself. For example the calling conventions seem to be a major bottleneck, now that the garbage collector has be sped up significantly 2) rakudo itself. For example the lazy list iteration code could certainly use some optimizations 3) For some features there's a mismatch between what parrot provides and what rakudo/Perl 6 needs, which makes rakudo go longer ways, which in turn implies slowness. A large example of this is that rakudo can't serialize all built-in types, but rather has to re-generate them at startup time - which of course is a costly affair. Cheers, Moritz -- --gh
Re: Fwd: FOSDEM - perl 6 critic
Am 22.02.2011 16:57, schrieb Gabor Szabo: He, as a sysadmin would like to do the small tasks in a relatively small language. He would like to make sure the modules/applications he will download and will have to support are in such a relatively small language. Whatever Perl 6 will turn out to be, it won't be a small language. If I have any influence, it will be a language that's pleasant to write and to use (and it largely is already), but it's not small. (Neither is Perl 5 a small language; Perl 6 is just another step larger). We can't be everybody's darling, as much as we would love to. I wouldn't be opposed to stripping down some parts of Perl 6 a bit, and make them available by loading (core) modules, but that's not the Perl 6 we are working on, but rather a different beast. Arne Wichmann wrote: My other gripe is that perl5 nowadays already is too big - it takes too much memory and time for small tasks. But that is only secondary. Here's the waterbed again: if we push it down on the one side and make it pleasant to use, other parts (resource usage) go up. That said, early Java compilers and VMs also sucked in terms of resource usage. We'll just have to see how well Perl 6 can be optimized, and if/when we can muster the resources to do it. Cheers, Moritz
Re: Fwd: FOSDEM - perl 6 critic
On Tue, Feb 22, 2011 at 10:46 AM, Moritz Lenz mor...@faui2k3.org wrote: We can't be everybody's darling, as much as we would love to. That's a fair statement, however do consider that perl5 is still a darling for many system administrators and command-line warriors who have long since left awk and sed far behind and never looked back. In this use case, perl5 still rules, and I would suspect that you wouldn't want to simply throw away these users. My (admittedly limited) experience with rakudo leads me to believe that the one-liner or power-tool usage we've come to expect with perl5 may simply not exist in perl6. [frank@zino00 ~]$ echo a b c | perl -lane 'print $F[1]' b [frank@zino00 ~]$ echo a b c | perl6 -lane 'print $F[1]' ===SORRY!=== Unable to open filehandle from path '-lane' [frank@zino00 ~]$ echo a b c | perl -pe 's/b/BEE/ if /^a/' a BEE c [frank@zino00 ~]$ echo a b c | perl6 -pe 's/b/BEE/ if /^a/' ===SORRY!=== Unable to open filehandle from path '-pe' These are the kinds of things I do every single day many times per day without a second thought. It would be a shame IMO if I couldn't continue to use perl6 in such a fashion. Just my own 2 cents. Thanks. --frank
Re: Fwd: FOSDEM - perl 6 critic
On Wed, Feb 23, 2011 at 3:10 AM, Frank S Fejes III fr...@fejes.net wrote: On Tue, Feb 22, 2011 at 10:46 AM, Moritz Lenz mor...@faui2k3.org wrote: We can't be everybody's darling, as much as we would love to. That's a fair statement, however ... Let me just link to the recent blog post of chromatic: http://www.modernperlbooks.com/mt/2011/02/unifying-the-two-worlds-of-perl-5.html and point out the link Aristotle posted: http://www.nntp.perl.org/group/perl.perl5.porters/2011/02/msg169141.html He enumerates some of the groups of Perl users that have distinctive usage patterns. [frank@zino00 ~]$ echo a b c | perl -lane 'print $F[1]' b [frank@zino00 ~]$ echo a b c | perl6 -lane 'print $F[1]' ===SORRY!=== Unable to open filehandle from path '-lane' [frank@zino00 ~]$ echo a b c | perl -pe 's/b/BEE/ if /^a/' a BEE c [frank@zino00 ~]$ echo a b c | perl6 -pe 's/b/BEE/ if /^a/' ===SORRY!=== Unable to open filehandle from path '-pe' I am not an expert on Perl 6 but I think the fact that those things you mentioned don't work on the command line of Rakudo is mainly as they were not *yet* implemented. They might not be fully specced yet either. See the draft version of the specification: http://perlcabal.org/syn/S19.html I am sure the Perl 6 developer would be happy if you could send them patches implementing some of those option or if at least you could write tests that check if the features work. regards Gabor
Re: Fwd: FOSDEM - perl 6 critic
On 02/23/2011 02:10 AM, Frank S Fejes III wrote: On Tue, Feb 22, 2011 at 10:46 AM, Moritz Lenz mor...@faui2k3.org wrote: We can't be everybody's darling, as much as we would love to. That's a fair statement, however do consider that perl5 is still a darling for many system administrators and command-line warriors who have long since left awk and sed far behind and never looked back. In this use case, perl5 still rules, and I would suspect that you wouldn't want to simply throw away these users. That's correct. But those users chose Perl because they get stuff done with it, not because it's a small language (and it's not). My (admittedly limited) experience with rakudo leads me to believe that the one-liner or power-tool usage we've come to expect with perl5 may simply not exist in perl6. As Gabor points out, they haven't been implement yet, but they will be. Cheers, Moritz