Hello.
I've been grappling with unit testing for a couple weeks now. I'm
a bourne shell scripter by day who is trying to expand my skill
set. I currently have a ruby script which I'm trying to create some
unit tests for. The unit tests have not been running and I've
finally traced the problem to this method. Specifically the "exit"
call. If I comment out the exit then my tests actually run.

# Cleanup any transient files and directories and then exit
def script_clean
FileUtils.rm_rf(WORKDIR) if File.exists?(WORKDIR)
# Everything is tidy we should leave
exit
end

The above method is called in 2 different scenarios. The first case
is during script execution if any Exceptions occur. Then the error
message is printed and the script_clean method is called to ensure the
process exits.

Here is an example of where the method is called

begin
File.open(ADMIN_FILE, "w+") {|f| f << ADMIN_CONTENTS }
rescue Exception => e
puts e ; script_clean
end

Additionally if the script runs through to the end without
encountering any errors then the last line is a call to the
script_clean method

script_clean

My tests are not triggering any Exceptions so it appears to be the
last line which calls "script_clean" to hamper any test execution.
Explicitly exiting when an uncorrectable error has happened is
something I commonly do in bourne shell scripts and it made sense to
carry over the practice in ruby. Should I be terminating premature
and normal script execution in some other fashion that is more
compatible with test/unit.

Here is an basic example demonstrating what I'm seeing.


------------example script to reproduce 'exit' call hindering unit
tests
require 'optparse'

# Argument handling
options = {}
OptionParser.new do |o|
o.on("-n") { |options[:noop]| }
o.on_tail("-h", "--", "Print usage.")
o.parse!(ARGV)
end
# Set PROTO constant based on command line arg
if options[:noop]
PROTO = true
else
PROTO = false
end

# Example method that exits
def clean
exit
end

# Call clean method if PROTO is true
if PROTO
clean
end

----------example unit test
# Load up required supporting files
require 'someapp'
require 'test/unit'

# Class to hold the config file tests
class CliTest < Test::Unit::TestCase
def test_dryrun
assert PROTO, true
end
end

-----------example execution where "exit" is not called
../test_someapp.rb
Loaded suite ./test_someapp
Started
F
Finished in 0.013832 seconds.

1) Failure:
test_dryrun(CliTest) [./test_someapp.rb:23]:
true.
<false> is not true.

1 tests, 1 assertions, 1 failures, 0 errors

----------example execution which calls "exit"
./test_someapp.rb -n
# echo $?
0


Any tips for a bourne shell convert are very much appreciated!
TIA. G

On Nov 19, 2007, at 17:05 , weathercoach@gmail.com wrote:

> If I comment out the exit then my tests actually run.
>
> # Cleanup any transient files and directories and then exit
> def script_clean
> FileUtils.rm_rf(WORKDIR) if File.exists?(WORKDIR)
> # Everything is tidy we should leave
> exit
> end

so don't call exit.

really.

since it is the last thing you're doing anyways, what is it doing for
you? Nothing.

if you HAVE to call exit (and you don't), then do this in your tests:

alias ld_exit :exit
def exit
# no
end

---

On second thought. Get rid of script_clean entirely. It is useless.
Use Tempfile instead and you'll be happier.

>
> On second thought. Get rid of script_clean entirely. It is useless.
> Use Tempfile instead and you'll be happier.


Ryan.
The script_clean method is also used when an exception is raised not
just as the last line of the script, if that was the case I would
agree with your recomendation to remove it. Some of the tests I had
planned on creating would trigger an exception to make sure my error
handling is correct. Part of the error handling is cleaning up any
transient files and exiting immediatly to ensure that the script does
not continue running and do something unintended.
If a call to "exit" impacts the execution of the unit tests I'm
trying to understand if I should be using a different approach.
Thanks for your suggestions!
G.

On Nov 19, 2007, at 18:45 , weathercoach@gmail.com wrote:

> If a call to "exit" impacts the execution of the unit tests I'm
> trying to understand if I should be using a different approach.

A call to exit impacts the execution of ruby... not just unit tests.

I still don't see why you need that method. Tempfile cleans up after
itself... that is what it is for.

2007/11/20, weathercoach@gmail.com <weathercoach@gmail.com>:
> Hello.
> I've been grappling with unit testing for a couple weeks now. I'm
> a bourne shell scripter by day who is trying to expand my skill
> set. I currently have a ruby script which I'm trying to create some
> unit tests for. The unit tests have not been running and I've
> finally traced the problem to this method. Specifically the "exit"
> call. If I comment out the exit then my tests actually run.
>
> # Cleanup any transient files and directories and then exit
> def script_clean
> FileUtils.rm_rf(WORKDIR) if File.exists?(WORKDIR)
> # Everything is tidy we should leave
> exit
> end
>
> The above method is called in 2 different scenarios. The first case
> is during script execution if any Exceptions occur. Then the error
> message is printed and the script_clean method is called to ensure the
> process exits.
>
> Here is an example of where the method is called
>
> begin
> File.open(ADMIN_FILE, "w+") {|f| f << ADMIN_CONTENTS }
> rescue Exception => e
> puts e ; script_clean
> end
>
> Additionally if the script runs through to the end without
> encountering any errors then the last line is a call to the
> script_clean method
>
> script_clean
>
> My tests are not triggering any Exceptions so it appears to be the
> last line which calls "script_clean" to hamper any test execution.
> Explicitly exiting when an uncorrectable error has happened is
> something I commonly do in bourne shell scripts and it made sense to
> carry over the practice in ruby.

I disagree. Ruby and Bourne Shell are two completely different
languages. Especially shell does not have exception handling. I
would especially not exit from unit tests but raise an exception.
Even if you use exit you can make sure your cleanup code is run by
placing it in END or a global "ensure" clause:

10:46:02 /cygdrive/c/SCMws/RKlemme/OPSC_Gold_bas_dev_R1.1.5_s11_pp/bas
$ ruby -e 'END{puts 1}; exit 2'
1
10:46:05 /cygdrive/c/SCMws/RKlemme/OPSC_Gold_bas_dev_R1.1.5_s11_pp/bas
$ ruby -e 'begin; exit 2;ensure puts 1; end'
1

> Should I be terminating premature
> and normal script execution in some other fashion that is more
> compatible with test/unit.

Yes, raise an exception. Btw, exit does actually throw an exception
but maybe that's not caught by the test framework.

$ ruby -e 'begin; exit 2; rescue Exception => e; p e, e.class,
e.class.ancestors; end'
#<SystemExit: exit>
SystemExit
[SystemExit, Exception, Object, Kernel]

You usually do not want your complete test suite to terminate when a
single test fails.

Kind regards

robert

--
use.inject do |as, often| as.you_can - without end

On 20.11.2007 03:43, weathercoach@gmail.com wrote:
>> On second thought. Get rid of script_clean entirely. It is useless.
>> Use Tempfile instead and you'll be happier.
>
>
> Ryan.
> The script_clean method is also used when an exception is raised not
> just as the last line of the script, if that was the case I would
> agree with your recomendation to remove it. Some of the tests I had
> planned on creating would trigger an exception to make sure my error
> handling is correct. Part of the error handling is cleaning up any
> transient files and exiting immediatly to ensure that the script does
> not continue running and do something unintended.

This sounds as if you did not yet get the hang of exception handling.
If an exception is so serious that the whole program cannot possibly
proceed then don't catch the exception (or catch and rethrow another
one). There are a lot mechanisms to ensure proper and robust cleanup,
e.g. "ensure" clauses, END, at_exit, ObjectSpace.define_finalizer and on
a higher level as Ryan pointed out: Tempfile, File.open with block ...

> If a call to "exit" impacts the execution of the unit tests I'm
> trying to understand if I should be using a different approach.

Yes, IMHO you should use a different approach - but not only to testing
but as well (and more important) to error handling. Please try to get
rid of your shell mindset when doing Ruby development.

Kind regards

robert