Re: Module::Build design plans
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Wed, 22 May 2002, Ken Williams wrote: I'm hoping to use YAML as the format for the metadata file, If anyone really thinks I really really have to use XML, speak up. Well... yeah... I can't see any reason for expecting users to learn a new markup format. XML is not ideal but it's widely known and *good enough*, just like Perl. I suppose that using somebody else's markup language (YAML) is better than making your own; then the wheel has been reinvented only once. But still, it seems like gratuitous incompatibility. If there are technical reasons to avoid XML in favour of YAML, then sure, go ahead. If it's just a question of taste, then I would suggest that following an established standard is more important when the two conflict. Just like you may not agree with perlstyle(1), but it works out best when everyone follows it. Although this is just my 2c. At this point it would almost be relevant for me to mention http://www.langdale.com.au/SOX/. - -- Ed Avis [EMAIL PROTECTED] Finger for PGP key -BEGIN PGP SIGNATURE- Version: GnuPG v1.0.6 (GNU/Linux) Comment: For info see http://www.gnupg.org iD8DBQE861hwIMp73jhGogoRAi3/AJ96492L1GWED+AnswFYDNDU2pL/mACgg7gT mlgHm3aMyQZOfISct1WSXZU= =J4QM -END PGP SIGNATURE-
Re: Module::Build design plans
On Wed, 22 May 2002, Edward Avis wrote: Well... yeah... I can't see any reason for expecting users to learn a new markup format. XML is not ideal but it's widely known and *good enough*, just like Perl. I suppose that using somebody else's markup language (YAML) is better than making your own; then the wheel has been reinvented only once. But still, it seems like gratuitous incompatibility. YAML ain't a markup language It's a data serialization language. And as long as no one intends that module authors actually _write_ it (by text editor) then I think it's just fine. If there are technical reasons to avoid XML in favour of YAML, then sure, go ahead. If it's just a question of taste, then I would suggest that following an established standard is more important when the two conflict. Just like you may not agree with perlstyle(1), but it works out best when everyone follows it. Although this is just my 2c. XML requires more tools than I think Ken wants to require. At a minimum you'd need XML::Parser and possible some XML writer module. YAML is a simple all-in-one solution. At this point it would almost be relevant for me to mention http://www.langdale.com.au/SOX/. Eek, Python as XML, or XML as Python, or somehting like that ! -dave /*== www.urth.org we await the New Sun ==*/
Re: Module::Build design plans
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Wed, 22 May 2002, Dave Rolsky wrote: Anyway, I think it's mostly moot. YAML is darn easy to use: print Store ($data_struct); my $data = Load($text); and that, along with human-readability and minimal requirements probably makes it a done deal. Data::Dumper is just as easy but as you point out it requires an eval() which allows all sorts of unpleasant surprises. It would be nice to have a 'safe' parser for Data::Dumper, because then you could use it to load and save data in the ideal format (in other words, that which is most familiar to Perl users). Unfortunately no such parser exists, and I haven't used Parse::RecDescent before so I am not able to hack one together in half an hour and attach it to this message :-(. So it looks like YAML may be the only option, at present. - -- Ed Avis [EMAIL PROTECTED] Finger for PGP key -BEGIN PGP SIGNATURE- Version: GnuPG v1.0.6 (GNU/Linux) Comment: For info see http://www.gnupg.org iD8DBQE867+zIMp73jhGogoRAjtsAJ4rx25BQwxwHPpbXuSocEmjntf7jACfa2xN LQTabyfGZv2Ca/O/C1J31OI= =wDnW -END PGP SIGNATURE-
Re: Module::Build design plans
On Wed, 22 May 2002, Michael G Schwern wrote:
I'm not quite sure I get this one. What's writing the YAML?
print $fh Store { version = $version, prereqs = \%prereqs };
All I meant was that no one will have to use a text editor to generate
actual YAML.
-dave
/*==
www.urth.org
we await the New Sun
==*/
Re: Module::Build design plans
On Wed, May 22, 2002 at 10:45:21AM -0500, Dave Rolsky wrote: On Wed, 22 May 2002, Edward Avis wrote: And as long as no one intends that module authors actually _write_ it (by text editor) then I think it's just fine. By that argument what is wrong with Storable? It's not human _readable_. Also, the default format isn't canonical. I *think* that the network order format is canonical (in that the source has explicit hacks that do *something* on Crays, apparently to read/write 32 bit quantities) but I wouldn't trust it utterly. And as well as the readability stuff, it has the big disadvantage that it can get mangled by any well-meaning ASCII file transfer. Nicholas Clark
Re: Module::Build design plans
ESEVERELYOVERLATINIZED On Wed, 22 May 2002 12:55:46 -0400, Michael G Schwern [EMAIL PROTECTED] wrote: So we basically still have the moral equivalent of a Makefile.PL, only it's run by the author before distribution to generate a static data file which is distributed avoiding the need to run or parse any code or download and unpack the whole distribution just to get information about the module?
Re: [PATCH MM 5.95_01] basic.t VMS dir depth revision
On Sun, May 19, 2002 at 04:31:02PM -0500, Craig A. Berry wrote: +$ IF F$SEARCH(t.DIR) .NES. +$ THEN +$! building CPAN version +$ BFD_TEST_ROOT = F$PARSE([.t]NO_CONCEAL)-.][00-][-].;+.] +$ ELSE +$! we're in the core +$ BFD_TEST_ROOT = F$PARSE(SYS$DISK:[]NO_CONCEAL)-.][00-][-].;+.] +$ ENDIF Could you change this to look for the PERL_CORE environment variable? -- This sig file temporarily out of order.
Re: MM 5.95_02 on VMS (was Re: [ANNOUNCE] ExtUtils::MakeMaker 5.93_01now backported to 5.005_03)
On Sat, May 18, 2002 at 07:49:52PM -0500, Craig A. Berry wrote: Perl 5.6.1 on OpenVMS Alpha 7.1 with just-downloaded MM snapshot now gives: t/command... # Failed test (t/command.t at line 53) # '00setup_dummy.%' # doesn't match '(?-xism:00setup_dummy.t)' # Looks like you failed 1 tests of 24. [looks like we need to use % for glob() and ? for a regex -- yuck] t/testlib... can't run t/testlib.t. invalid argument FAILED before any test output arrived [very odd, runs fine by itself. perhaps something about -T] Try the snapshot. It should work but you'll have to upgrade Test::Harness. -- This sig file temporarily out of order.
Re: Module::Build design plans
On Thursday, May 23, 2002, at 02:38 AM, Michael G Schwern wrote:
On Wed, May 22, 2002 at 10:31:27AM -0500, Dave Rolsky wrote:
YAML ain't a markup language
It's a data serialization language.
And as long as no one intends that module authors actually
_write_ it (by
text editor) then I think it's just fine.
I'm not quite sure I get this one. What's writing the YAML?
It would be written by YAML as part of the 'make dist' ('Build
dist', actually), from options given in the Build.PL file.
-Ken
Re: Module::Build design plans
On Thursday, May 23, 2002, at 02:55 AM, Michael G Schwern wrote: So we basically still have the moral equivalent of a Makefile.PL, only it's run by the author before distribution to generate a static data file which is distributed avoiding the need to run or parse any code or download and unpack the whole distribution just to get information about the module? Yup, nicely put allinonesentence. ;-) -Ken
