Practices


#1

Is there anything of a best practices on how to use your own modules?

I’ve been doing a LOT of this in perl and have just started to play with
it in Ruby.

Some things I am transistioning from are:
use lib – how do I specify an alternate location for files?

modules/object/files:
It may be bad form to put more than on class in a single file, I know it
is in
Perl. But is there a reason not to? I have some classes that are very
tightly
related in inheritance and they are very small. I don’t see much
advantage to
ever physically breaking them apart.


#2

I have been wondering the same. By looking at libraries out there, it
looks like there isn’t a standard view about this.

Does anybody have a pointer to conventions to structure
directories/files/modules
and classes?


#3

2006/4/19, Tom A. removed_email_address@domain.invalid:

Is there anything of a best practices on how to use your own modules?

I’ve been doing a LOT of this in perl and have just started to play with it in Ruby.

Some things I am transistioning from are:
use lib – how do I specify an alternate location for files?

There is an env var, I believe it’s RUBY_LIB.

modules/object/files:
It may be bad form to put more than on class in a single file, I know it is in
Perl. But is there a reason not to? I have some classes that are very tightly
related in inheritance and they are very small. I don’t see much advantage to
ever physically breaking them apart.

Then don’t do it. It’s fine to have them in the same file.

Kind regards

robert


#4

Jonas P. wrote:

Le 19 avr. 06 à 11:09, Robert K. a écrit :

There is an env var, I believe it’s RUBY_LIB.

It is RUBYLIB

Thank you. I found more information on it.
Seems that it will work but it has limitations, like safe mode.

If I have a safemode > 0 and files in ~/lib/ how would I use them?
I can’t edit the system environment in this case.
Would I convert the ~/ to the fullpath and use require:

require ‘/home/username/lib/noclass’

?


#5

Le 19 avr. 06 à 11:09, Robert K. a écrit :

There is an env var, I believe it’s RUBY_LIB.

It is RUBYLIB

Cheers,
zimbatm


#6

On 4/19/06, Tom A. removed_email_address@domain.invalid wrote:

If I have a safemode > 0 and files in ~/lib/ how would I use them?
I can’t edit the system environment in this case.
Would I convert the ~/ to the fullpath and use require:

require ‘/home/username/lib/noclass’

?

I think a fairly common idiom in cases like this (one I have used
myself) is to add to the LOAD_PATH within your script (at the top
before your requires):

$LOAD_PATH.unshift ‘~/lib’

Then require will check in ~/lib for scripts before checking the other
default locations.

Ryan


#7

On 4/18/06, Tom A. removed_email_address@domain.invalid wrote:

modules/object/files:
It may be bad form to put more than on class in a single file, I know it is in
Perl. But is there a reason not to? I have some classes that are very tightly
related in inheritance and they are very small. I don’t see much advantage to
ever physically breaking them apart.

Will you ever load them without each other? If so, then split them. If
not, don’t.

-austin


#8

If the safe mode is a problem, do all your require’ing before changing
the safe mode (I think). You might also have an unsafe program that
does the require’ing, then includes and runs the safe part.

For non-trivial programs, I have begun using a structure as follows. I
picked up the bin/ doc/ test/ somewhere (thanks whomever you are), but
can’t remember where, and added to it.

labeler/
bin/ # “Binaries,” i.e. files that are run directly
lbl.rb
config/ # config files.
config.yaml
label_defs.yaml
data/ # Any data needed to setup/run the system
doc/ # RDoc and other documentation
classes
files

email-from-blah-blah.txt
some-ref.pdf
lib/ # rb’s go here
entity.rb
label.rb

release/ # Files needed to build a release
build.rb # of the system go here.
labeler-1.1.5.zip # zip created to send customer
# *.zip is ignored by svn here.
test/ # unit tests go here
data/ # it would be better if my test data
# were in a dir like this. currently
# it’s all just glommed into test/.
tc_lbl.rb # tc_ = test case
test_labeler.rb

ts_labeler.rb # a test suite
vendor/
extensions/ # my own extensions to Ruby core classes
integer_ext.rb # adds #is_odd?, #is_even?
regexp_ext.rb # adds #+ method.
serial/ # a serial library
vendor.rb # see below

In my main program, bin/lbl.rb, I have the following:

Add …/lib and …/vendor to the load path.

$:.unshift File.expand_path(File.join(File.dirname(FILE), “…”,
“lib”))
$:.unshift File.expand_path(File.join(File.dirname(FILE), “…”,
“vendor”))
require ‘vendor’
require ‘optparse’

require ‘label_printer’

Location of the config file.

CONFIG_FILE = File.join(File.dirname(FILE), “…”, “config”,
“config.yaml”)

config = YAML.load_file(filename) # load the config data

####################

vendor/vendor.rb is the following:

#!/usr/bin/env ruby

vendor

Created by John J. on 2006-03-31.

Copyright © 2006 John J. Software, LLC, All rights reserved.

Prepend all vendor subdirectories to the load path.

dir=Dir.new(File.dirname(FILE))
dir = dir.to_a
not_dots = dir.delete_if { |d| d =~ /^./ }
not_files = not_dots.delete_if { |d|
to_check = File.join(File.dirname(FILE), d)
File.file?(to_check)
}
not_files.each { |a_dir|
to_add = File.join(File.dirname(FILE), a_dir)
to_add = File.expand_path(to_add)
$:.unshift to_add
}

Which prepends all subdirectories of vendor onto the load path.
###################

I added an option to the main program (–test) that will run all the
unit tests:

opts = OptionParser.new

opts.on("–test", “Run unit tests on this program.”) {
Dir.chdir(‘test’)
system(‘ruby ts_labeler.rb’)
exit(1)
}

This is so the customer (or I) can do:
lbl --test
at my direction to test the system.

################

Since I deploy this system on Windows, I have the obligatory lbl.bat
file in the system’s root:
@echo off

rem Use OLD_PATH to keep paths found from accumulating in PATH.
SET OLD_PATH=%PATH%

rem Prefered path is last.
IF EXIST c:\djgpp\bin\ruby.exe SET PATH=c:\djgpp\bin;%OLD_PATH%
IF EXIST c:\usr\local\bin\ruby.exe SET PATH=c:\usr\local\bin;%OLD_PATH%
IF EXIST c:\ruby\bin\ruby.exe SET PATH=c:\ruby\bin;%OLD_PATH%

ruby bin/%0.rb %*
###################

All the above (files, dirs, data, etc.) is kept in an SVN repository.
Everything in the vendor folder is external, meaning they are housed
elsewhere, and brought into vendor on this system and others I write.

For small systems or utilities for my own use, with a couple of files,
I just put everything in one directory.

As far as dividing classes into files, go with your gut. I put like
things together. There is no need in having thirty’leven (colloquialism
used in the Southeastern United States meaning ‘a lot of’, see also
‘dollar two ninety eight’) files just so each class is in it’s own
file. Writing the unit tests seems to help with deciding how to group
things too. Unit testing also helps you divide functions and methods
into manageable chunks.

I highly recommend using Subversion and unit tests (if you couldn’t
tell). There is a learning curve, but both are indispensable.

Hope this helps!

Regards,
JJ

On 18-Apr-2006, at 20:17, Tom A. wrote:

it is in Perl. But is there a reason not to? I have some classes
that are very tightly related in inheritance and they are very small.
I don’t see much advantage to ever physically breaking them apart.


Help everyone. If you can’t do that, then at least be nice.


#9

Spectacular! Thanks.