* We were converting Exabytes to bytes as Petabytes.
* Updated tests to cover ever unit.
* Added note that we're going by the old, inaccurate definitions of
Kilobytes, Megabytes, etc, in that we treat them as powers of 2.
We need to use
unless value.is_a?(String) || value.is_a?(Array)
rather than
klass = value.class
unless [String, Array].include?(klass)
because the klass version enforces type checking which is too strict, and does
not allow us to accept objects wich have extended String (or Array).
For example, generate() function now returns Puppet::Util::Execution::ProcessOutput
which is just a very simple extension of String. While this in it's self was
not intentional (PUP-2306) it is not unreasonable to cope with objects which
extend Strings
simplecov 0.9 dropped ruby 1.8 support, and stdlib is one of the oddball
modules that uses it. So we could probably just remove it and be okay.
(cherry picked from commit a7c129b22d)
It was discovered that the concat array modifies the arrays passed to it
as an argument as a side effect. This test will ensure that doesn't
happen again.
- ensure_packages fails because Error: Sun packages must specify a package source
- ensure_resource fails for the same reason
- get_module_path fails because the modulepath is different
- has_interface_with fails because the interface is lo0 not lo
This conversion is done by Transpec 2.2.1 with the following command:
transpec spec/functions
* 345 conversions
from: obj.should
to: expect(obj).to
* 122 conversions
from: == expected
to: eq(expected)
* 85 conversions
from: lambda { }.should
to: expect { }.to
* 22 conversions
from: be_true
to: be_truthy
* 16 conversions
from: be_false
to: be_falsey
* 11 conversions
from: pending
to: skip
* 9 conversions
from: it { should ... }
to: it { is_expected.to ... }
* 5 conversions
from: =~ [1, 2]
to: match_array([1, 2])
* 2 conversions
from: =~ /pattern/
to: match(/pattern/)
* 2 conversions
from: obj.should_not
to: expect(obj).not_to
For more details: https://github.com/yujinakayama/transpec#supported-conversions
This conversion is done by Transpec 2.2.1 with the following command:
transpec spec/unit
* 53 conversions
from: obj.should
to: expect(obj).to
* 19 conversions
from: == expected
to: eq(expected)
* 5 conversions
from: lambda { }.should
to: expect { }.to
* 2 conversions
from: be_true
to: be_truthy
For more details: https://github.com/yujinakayama/transpec#supported-conversions
file_line supports adding lines after a match, but there are use cases when
having "before" would be useful. For example, in Debian-based OS's, the last
line of /etc/rc.local is "exit 0" it's an incredible pain to deal with
that scenario today.
This commit adds a 'before' parameter to the file_line type, and implements
it for the ruby provider.
rspec-puppet matchers are defined for tests which exist in
spec/functions, but the function unit tests lived in
spec/unit/puppet/parser/functions. This moves them to the correct place
for using rspec-puppet
Previously this was incorrectly handling facts that were of the form
foo=1+1=2 due to the ='s in the actual fact contents. Fix this and
add tests to try and prevent regressions.
This test attempts to emulate various versions of facter, but is
still dependent on the version of facter it is running against. The
immediate symptom was that the test breaks with facter 2.0.1 because
it adds another external facts search directory.
I tried a couple ways to stub this but allowing it to pretend
to run against one set of facters, while actually running against
one real facter (which might itself be one of several versions)
eluded me.
So this patch just removes the test.
Without this patch one can not specify package resource specific
parameters. All the ensure_packages() function does it makes sure
the named packages are installed. This patch allows one to pass
default as a second argument and allow greater flexibility on
packages installations.
Use case like the following are now possible :
* ensure_packages(['r10k', 'serverspec'], {'provider' => 'gem'})
* ensure_packages(['ntp'], {'require' => 'Exec[foobar]'})
This work updates a number of Gems to the latest versions (rspec,
rspec-puppet), and updates and tweaks a bunch of tests to work
with the updated gems.
Instead of modifying the first paramater of deep_merge due to the
use of the merge! function, instead use merge to return a copy of
the merged object. This allows one to continue to use the original
first parameter after the call to deep_merge.
I expect a function called "is_numeric" or "is_integer" to check if a
variable is an integer or a number even if the variable passed by isn't
a string nor a number at all. Otherwise we should call them
is_string_a_number and is_string_an_integer and we have then to remove
the check for .is_a?(Number) and .is_a?(FixNum)
now checking also if it is a hex or octal number
improved/corrected checking for integer
* checking against Integer instead of Fixnum so that
also Bignum is matching
* now .is_a? Integer is done first so this is quiet fast
Now many types of numerics are recognized.
1. Float/Integer values (signed or unsigned, with exponent or without)
2. octal and hex check
3. except hex numbers and the "0." in a float lower than 1 can be prefixed
with a '0'.
whitespaces shouldn't be allowed as prefix/suffix
string representation of numbers should not contain any type of
whitespace.. the user is responsible to clean a string before checking
it..
fix documentation and added more checks
tried to be 99.9% backward compatible
* for now the decission is post poned if hex and octal numbers
should be allowed or not (is_numeric)
* native Bignum is now also a valid integer class
fix problem with old 1.8 ruby and Hash.to_s/Array.to_s
In ruby < 1.9 array and hashes would be recognized as numeric
if they have a special format:
1.8:
[1,2,3,4].to_s = "1234"
{1=>2}.to_s = "12"
1.9:
[1,2,3,4].to_s = "[1, 2, 3, 4]"
{1=>2}.to_s = "{1=>2}"
The previous behavior of the tests checked the behavior of the
underlying functions library when called with no arguments; this commit
updates the tests to conform to the functions API and test what happens
when a function is called with no args.
Facter 1.7.4 changed how it decides on what directory to look in for
facts.d based on the user it is running as. This stubs out that bit of
code to make it think it is running as root.
Issue #20200 notes that the merge function does not
support nested hashes.
To prevent unintended side effects with changing merge,
add a deep_merge function instead.
The issue #20681 describe the error of delete() function
removing the elements from the origin array/hash/string.
This issue affected other delete functions. Because
ruby delete and delete_if functions make destructive
changes to the origin array/hash.
The delete_undef_values removed elements from the
origin array/hash and this is not the desired behaviour.
To solve this, we should dup or clone the array/hash
before using the delete or delete_if ruby functions.
We should also check if args[0] is not nil before using
dup, since dup on nil raises exception.
This fix the problem and add unit tests, so we could
enforce this behaviour and prevent regressions.
As stated on the issue #16498, it would be great to have unit tests
for all the functions.
Function loadyaml was missing a unit test. This commit added the
unit test to loadyaml function.
The issue #20681 describe the error of delete() function
removing the elements from the origin array/hash/string.
This issue affected the other delete functions.
The delete_at function is not afected by this bug, but
it did not had the unit test to check against it.
I had added the unit test so we could prevent regressions
on the future and also have better test coverage.
The issue #20681 describe the error of delete() function
removing the elements from the origin array/hash/string.
This issue affected other delete functions. Because
ruby delete and delete_if functions make destructive
changes to the origin array/hash.
The delete_undef_values removed elements from the
origin hash and this is not the desired behaviour.
To solve this, we should dup or clone the hash
before using the delete or delete_if ruby functions.
This fix the problem and add unit tests, so we could
enforce this behaviour and prevent regressions.
The setup: list with 3 elements, delete one:
$test_list = [‘a’, ‘b’, ‘c’]
$test_deleted = delete($test_list, ‘a’)
Print out the elements in ‘test_deleted’:
notify { ‘group_output2’: withpath => true, name => “$cfeng::test_deleted”, }
Notice: /Stage[main]/Syslog/Notify[group_output2]/message: bc
Good! Run-on output shows that ‘a’ was deleted
Print out the elements in ‘test_list’:
notify { ‘group_output1’: withpath => true, name => “$cfeng::test_list”, }
Notice: /Stage[main]/Syslog/Notify[group_output1]/message: bc
WHAT!? 'a' was deleted from ‘test_list’ as well! Expected abc as output!
This behaviour is confirmed for string, hash and array.
This is fixed on this commit, I had added two spec tests to cover that cases.
bug #20681 spec test for delete() function.
I had forgot in the last commit the spec test for hash in the
delete function.
bug # 20681 delete() function change aproach.
Instead of rejecting elements from the original list, we use
collection = arguments[0].dup .
then latter we could continue to use delete and gsub! on collection
without impact on original argument.
this is a better solution than the previous one, and works on ruby
1.8.7, 1.9.3 and 2.0.0.
The previous solution does not work on ruby 1.8.7.
delete function remove typo whitespace.
fix typo whitespaces.
When pick function fail return a better error message like
the other stdlib functions, indicating that the error
is on function pick.
This would help people that see the error to identity it is
related to a incorrect use of stdlib function pick, instead of having
to grep all puppet libraries and manifests source for the old message.
I had also changed the spec test.
pick function change spec as suggested GH-179
Fix the spec test to use expect {}.to instead of lambda {}.should
as explained by Adrienthebo.
"Using expect { }.to is preferred over lambda { }.should.
In addition it's best practice to do a string match against the
error message to ensure that we're catching the right error,
instead of any error of the right type."
Also fixed a typo on the error message, it was missing one space.
pick function stylish fix as suggested on GH179
Without this patch the implementation of the file_line provider is a bit
convoluted with respect to the newly introduced "after" parameter.
This patch addresses the problem by separating out the concerns of each
case into their own methods of handling the behavior with the match
parameter, handling the behavior with the after parameter, or simply
appending the line.
When adding new lines to a file the 'after' option can be useful
when you need to insert file lines into the middle of a file.
This is particularly helpful when using file_line with sectioned
config files.
NOTE: the after option only works when adding new lines. If you are
updating an existing (matched) line it will simply modify it in place.
This assumes it was in the right place to begin with.
Without this patch the stdlib spec tests are failing against recent
versions of Puppet. The root cause of this problem is a change in the
behavior of create_resources in Puppet 6baa57b. The change in behavior
caused the :name key to be omitted from the hash returned by
Puppet::Parser::Resource#to_hash which in turn is causing the test
failure.
This patch addresses the problem by updating the test to match the
description of the example. Only the attribute :ensure is checked
instead of the full hash itself.
An optional third parameter can be given a min length. The function
then only passes successfully, if all strings are in the range
min_length <= string <= max_length
update and fix function and unit tests
the check for the minlength has to be written differently
because 0 values should be possible. We now check
a) if the input is convertible, and throw a ParseError and
b) if the input .is_a?(Numeric) and ask for a positive number
it's not as clean as for maxlength, but keeps a similar behaviour
refined the error checking for the min length
try to convert to Integer(args[2]) and fail,
if it's not possible
changed the tests accordingly to the new parameter checking
Without this commit the file_line type will outright fail if multiple
lines match the given regex. This commit allows the file_line type and
provider to optionally match and modify all matching lines.
Changeset rebased into a single commit by Adrien Thebo <adrien@puppetlabs.com>
Without this patch Puppet Enterprise users who install the most recent
version of stdlib lose the ability to resolve certain facts critical to
the operation of Puppet Enterprise. These facts are defined externally
in the file
`/etc/puppetlabs/facter/facts.d/puppet_enterprise_installer.txt`.
As an example, Puppet Enterprise catalogs fail to compile if the
`fact_stomp_server`, and `fact_stomp_port` facts are not defined.
`facter_dot_d` was removed from stdlib version 4 because Facter version
1.7 now supports external facts defined in
`/etc/puppetlabs/facter/facts.d/puppet_enterprise_installer.txt`.
Puppet Enterprise does not yet include Facter 1.7, however. The most
recent PE release, 2.8.1, includes Facter 1.6.17. With this version of
Facter, users who replace the version of stdlib that ships with PE with
the most recent version from the Forge will lose the ability to resolve
facts from
`/etc/puppetlabs/facter/facts.d/puppet_enterprise_installer.txt`.
This patch addresses the problem by detecting if Facter version < 1.7 is
loaded. If so, then the facter_dot_d.rb facts will be defined using the
stdlib custom fact. If Facter >= 1.7 is being used then stdlib will not
define external facts.
This patch allows an array of resource titles to be passed into
the ensure_resource function. Each item in the array will be
checked for existence and will be created if it doesn't already
exist.
Without this patch we don't have a spec test to make sure the anchor
resource type passes notify events along. This patch addresses the
problem by using RSpec Puppet to build a catalog, apply the catalog,
then pull a resource out of the transaction and make sure the resource
was triggered.
Without this patch the expected behavior of the count() function when
dealing with an out of bound array index and with a hash key that does
not exist is implicitly encoded in the spec examples. This is a problem
because the expected behavior is not clear for something similar to the
following example:
node default {
$ary = [ 1, 2, 3 ]
$ary_undef = $ary[100]
$hsh = { 'one' => 1 }
$hsh_undef = $hsh['dne']
$count = count(['hi', $ary_undef, $hsh_undef])
notice "Count is ${count}"
}
This patch addresses the problem by making the expected behavior
explicit in the examples.
This change is to implement a new function "any2array", which will take any
argument or arguments and create an array which contains it. If the argument
is a single array then it will be returned as-is. If the argument is a single
hash then it will be converted into an array. Otherwise (if there are more than
one argument, or the only argument is not an array or a hash) the function will
return an array containing all the arguments.
This is a bit more heavy-handed than I might like, but it does appear to
do the right things:
* accepts numeric input appropriately, truncating floats
* matches string input against a regex, then coerces number-looking
strings to int
* makes a best effort to coerce anything else to a string, then subjects
it to the same treatment
* raises an error in the event of incorrect number of arguments or
non-number-looking strings
I've also included some additional unit tests.
No more coercing to String and regex matching. Instead, we now coerce
to Integer at the beginning or raise an error if we cannot coerce to
Integer.
A consequence of this change is that the function will now accept
blatantly non-numeric strings as input, and return false. This seems a
bit goofy to me, but it's how String#to_i works. If we really don't
like this, then I'm open to suggestions.
Puppet passes numbers as String to functions, but it makes more sense to
compare them as Numeric.
But sometimes Puppet passes them as the wrong type, see:
https://projects.puppetlabs.com/issues/19812
This function provides a simple wrapper around
Puppet::Parser::Functions.function for access within Puppet manifests.
This will allow users to check whether or not a plugin or functionality
such as hiera is installed on the server.
This reverts commit f7a18189ec, reversing
changes made to 36a7b29630.
I'm reverting this change because of concerns raised by Peter Meier that
it duplicates the "in" operator in the DSL. The "in" operator is new
information that I did not posses when I made the decision to merge.
Because of this new information I'm un-merging and continuing the
discussion in the comments of
https://projects.puppetlabs.com/issues/19272
Reference: GH-130
It is exceptionally difficult to determine if an array contains an element matching a specific value without an iteration or loop construct.
This function is the Puppet equivalent of Array.includes?(foo) in Ruby. The implementation is a verbatim copy of has_key() with the minor modifications needed to support arrays instead of hashes.
Without this patch applied there is no easy way to append one array to
another. This is a problem because it is often desirable to join two
arrays without flattening the contents into a single, one dimensional
array.
This patch addresses the problem by adding a `concat()` function which
takes two arguments. The arguments will be concatenated together and a
new array returned to the caller.
Reviewed-by: Jeff McCune <jeff@puppetlabs.com>
As far as i know there's no other puppet-dsl-like way to get parameter of
defined resource, so that's why i implemented getparam function, which takes
resource reference and parameter name and returns parameter value.
Here's another example why this function is really useful:
define config($path, $config_param1, $config_param2) { }
define example_resource($config) {
$path = getparam($config, "path")
notice("Path is $path")
}
define example_resource2($example_resource, $config = getparam($example_resource, "config")) {
$config_param1 = getparam($config, "config_param1")
notice("Config parameter is $config_param1")
}
define example_resource3($example_resource, $config = getparam($example_resource, "config")) {
$config_param2 = getparam($config, "config_param2")
notice("Config parameter is $config_param2")
}
class test_getparam {
config { "config_instance":
path => "/some/config/path",
config_param1 => "someconfigtext1",
config_param2 => "someconfigtext2",
}
example_resource { "example_resource_instance":
config => Config["config_instance"]
}
example_resource2 { "example_resource_instance":
example_resource => Example_resource["example_resource_instance"]
}
example_resource3 { "example_resource_instance":
example_resource => Example_resource2["example_resource_instance"]
}
}
class { "test_getparam": }
* 4.x:
Add test/validation for is_float if created from an arithmetical operation
Add test/validation for is_integer if created from an arithmetical operation
Add test/validation for is_numeric if created from an arithmetical operation
* 2.x:
(Maint) Add spec/functions to rake test task
Add example behaviors for ensure_packages() function
Add an ensure_packages function.
Conflicts:
Rakefile
Without this patch the ensure_packages() function has no rspec behavior
examples. This patch fixes the problem by filling out a spec file with
expected behaviors I could think of.
Without this patch applied the file_line autorequire examples are
failing. This is a problem because the failures are false positives and
should be passing given the implementation.
This patch fixes the problem by changing the examples to directly test
the existence of the relationship by finding it in the list of
autorequire relationships.
If we manage a file we edit with file_line, it should be autorequired by
file_line. Without this patch applied the relationship is not
automatically setup and the user is forced to manually manage the
relationship.
This commit adds a function that joins each of a hash's keys with that
key's corresponding value, separated by a separator string. The
arguments are a hash and separator string. The return value is an
array of joined key/value pairs.
Previous to this commit, the delete function only acted on
arrays. This commit adds the same functionality for hashes and strings
in the obvious way: delete(h, k) would delete the k key from the h
hash and delete(s, sub) would delete all instances of the sub
substring from the s string.
This function is similar to a coalesce function in SQL in that it will
return
the first value in a list of values that is not undefined or an empty
string
(two things in Puppet that will return a boolean false value).
Typically,
this function is used to check for a value in the Puppet
Dashboard/Enterprise
Console, and failover to a default value like the following:
$real_jenkins_version = pick($::jenkins_version, '1.449')
The value of $real_jenkins_version will first look for a top-scope
variable
called 'jenkins_version' (note that parameters set in the Puppet
Dashboard/
Enterprise Console are brought into Puppet as top-scope variables), and,
failing that, will use a default value of 1.449.
If one wishes to test if a host has a particular IP address (such as a floating
virtual address) or has an interface on a particular network (such as a
secondary management network), the facts that provide this information are
difficult to use within Puppet.
This patch addresses these needs by implementing functions
‘has_ip_address(value)’ and ‘has_ip_network(value)’. These functions look
through all interfaces for ipaddress_<interface> and network_<interface>
(respectively) having the requested <value>.
These functions are implemented on top of a lower-level predicate
function, ‘has_interface_with(kind, value)’, which iterates through the
interfaces in the ‘interfaces’ fact and checks the facts <kind>_<interface>
looking for <value>.
Additionally, the existence of a particular named interface can be checked for
by calling with only a single argument: has_interface_with(interface).
A Boolean is returned in all cases.
This reverts commit 9e8c60a8b7.
This was an error on my part. 3.1.x descends from 3.0.x _and_ 2.5.x, but 3.0.x does
not descend from 2.5.x. I should not have merged 2.5.x into 3.0.x,
instead I should have merged 2.5.x into 3.1.x skipping over the 3.0.x
merge up.
I'm slowly starting to understand the implications of semver on our
branching strategy... =)
This reverts commit c8c7afc7a0, reversing
changes made to 03ec16e291.
This was a mistake on my part. 2.5 is a minor release and minor
releases should never be merged into patch level branches. 2.5.x should
have been merged into 3.1.x instead.
Without this patch stdlib tests fail against Facter 2.x and master but
not 1.6.x.
This patch fixes the problem by initializing the example group
differently depending on the version of Facter integrating into the
system. The adjusted methods are:
1.x - Facter.collection.loader.load
2.x - Facter.collection.load
The collection actually implements the load method in both version, it's
simply marked as private in 1.x.
* 2.5.x:
Update Modulefile, CHANGELOG for 2.5.0
Explicitly load functions used by ensure_resource
re-formatting
Add better docs about duplicate resource failures
Handle undef for parameter argument
Add function ensure_resource and defined_with_params
Revert "Revert "Merge branch 'haus-add_pe_facts_to_stdlib' into 2.4.x""
Revert "Merge branch 'haus-add_pe_facts_to_stdlib' into 2.4.x"
Prevent undefined method `split' for nil:NilClass with pe_foo_version facts
(maint) Clear all facts before each example
Add spec tests for pe_version facts
Add PE facts to stdlib
Conflicts:
CHANGELOG
Modulefile
spec/spec_helper.rb
This commit adds 2 new functions with unit tests.
defined_with_params works similarily to puppet's defined
function, except it allows you to also specify a hash of
params. defined_with_params will return true if a resource
also exists that matches the specified type/title (just like
with defined) as well as all of the specified params.
ensure_resource is a function that basically combines defined_with_params
with create_resources to conditionally create resources only if the
specified resource (title, type, params) does not already exist.
These functions are created to serve as an alternative to using
defined as follows:
if ! defined(Package['some_package']) {
package { 'some_package': ensure => present,
}
The issue with this usage is that there is no guarentee about
what parameters were set in the previous definition of the package
that made its way into the catalog.
ensure_resource could be used instead, as:
ensure_resource('package', 'some_package', { 'ensure' => 'present' })
This will creat the package resources only if another resource does
not exist with the specified parameters.
This reverts commit d6d23b495c.
This backwards-compatible additional functionality is targeted at the
next minor release. There are already backwards-incompatible changes in
the master branch so we need to establish a new minor branch.
This reverts commit 74e6411157, reversing
changes made to 417d219aa6.
Here's why:
Actually... I just screwed this up.
I merged this new fact into 2.4.x but it's not fixing any bug. It's adding a
new fact, so this should go into master and we should release 2.5 since this is
new, backwards-compatible functionality.
