App::TestOnTap - Test driver
Run a suite of test 'executables' in a directory tree, summarize their TAP output and optionally save the aggregated results for later processing.
Version 0.006
# Normally run 'testontap' from a shell prompt: # testontap --jobs 10 path/to/suite/root
NOTE: THIS IS CURRENTLY CONSIDERED TO BE IN BETA STATE. IT MAY CHANGE IN INCOMPATIBLE WAYS BEFORE THE 1.000 RELEASE.
This package is an application, generally intended to be run from a command line.
The purpose of the application is to act as a "driver", in terms of automatically executing arbitrary test scripts/executables (the "test suite") which in turn perform tests against whatever they're written to test (the "system under test", "SUT"). As such, the application has no idea what the tests actually do, it is only concerned about receiving TAP data back.
To help make the test suite comparatively easy to write and maintain, tests will be autodiscovered in it, based on a test suite configuration, and also, using rules in this configuration, enable control of the execution order of the tests.
Finally, it can optionally aggregate all TAP data into a defined format (e.g. a directory tree or directly to an archive) for later evaluation or other handling elsewhere (e.g. into a database with a custom UI on top).
So, summarizing this into a simple list of general requirements:
Scalable and easy to create small or large test suites.
Test suites could be small, developer oriented with just a few basic smoke-test scripts for quick system- or unit-testing during development, all the way to an enormous number of executables in a test suite for lengthy and in depth system-testing of a deployed system. Therefore the test suite should be easy to set up as an arbitrarily complex directory structure.
Preferably, writing a new test should require focusing almost entirely on the complexities of the test itself (which can certainly have its own challenges!) rather than how it should fit into the suite, at least at first. Thus, it's beneficial if there by default is no particular administration around a given test - just create the test script and it's automatically found and run.
Allow tests to run in a defined order, including when running in parallel.
In many cases, the time it takes to run a test suite is pretty much irrelevant, perhaps because it's fairly seldomly used and run only at night or simply because the tests are few and/or very quick. In these situations, it's ok to have the test driver run the tests in basically any order (currently, simple alphabetic order is used).
However, it may turn out that tests can be parallelized. In the best case, no test is dependent on another, so any amount of parallelization will work (given what a specific host can comfortably handle). In more complex cases, tests may depend on others in various ways so a mechanism is needed to express such dependencies, allowing the driver to schedule test scripts appropriately whether the tests are run one at a time or using 'n' possible parallel slots.
Handle result aggregation for entire test suite, with optional persistence of this data.
In a typical smoke-test test suite used during development, it's probably enough to just rely on eyeballing the test result, so it's enough with simple console based output on which tests worked and which ones failed (including which individual micro-tests failed).
For something like automated daily system tests however, it's typically important to retain the results for perusal later. In more advanced scenarios, such persisted results could be used to build a history of test results, for example measuring trends over the past months, creating graphs of data or similar. For this reason, it is useful if the test results can be packaged in a practical format, e.g. as an archive.
Very loose coupling between this application and test suite, system under test and other variables.
The implementation of the driver should have very little to do with how the tests or the SUT is implemented, or what or how they perform their stuff.
Language and platform agnostic.
Assuming the SUT and/or tests (perhaps exercising the SUT in a remote manner) can run on several platforms, the test driver should also run on the same platforms. Further, consistent with the previous point about loose coupling, the tests and SUT should be possible to implement in whatever language is most convenient.
This application tries to satisfy the above requirements. It provides one cornerstone of what I see as the minimum triplet of independent parts needed to run tests:
Thus, this application provides the first building block in this list. As discussed, it is not really concerned with what the tested product is, nor how the tests exercise it, only how the tests report results.
Each of these parts generally has its own CM and lifecycle need, and may each contain problems that need to be fixed - i.e. when a test fails, it's not necessarily the product that is faulty, the tests or the tool may be too, and so any of them may need adjustment without touching the others. Thus, it for example allows us to use any given version of tests and run them against a range of product versions, say, for finding regression problems.
In practice you may find you need even more parts to complete a full test environment, so unless you hardcode such things in the tests themselves, this application is open enough for you to pass in a number of variable parts - examples would be selecting several different JDK versions, or perhaps different incarnations of a database.
The core mechanism is to use TAP (e.g. "Test Anything Protocol" as defined by http://testanything.org) as the coupling between test executables and the application as the test driver, e.g. very similar to when writing regular Perl module tests. To implement this, the standard "Test::*" and "TAP::*" families of test modules are used, but with the necessary packaging to provide for the other requirements. Since the interface is 'TAP', any language can be used to write the tests.
The definition of a "test suite" includes a configuration file which allows the test suite to directly define which parts of the directory tree should not be scanned for possible tests, how to map "test executables" to how to execute them, as well as a dependency description mechanism for describing required dependencies between tests.
A test suite is basically very simple: it's a directory. To be minimally useful, it needs to contain a single file named with an extension defined in the default "exec mapping" (see more about defaults in configs execmap section).
So, assuming this:
foo/ -- bar/ -- -- simple_suite/ -- -- -- mytest.pl
Nothing more is necessary. By running "testontap foo/bar/simple_suite", it will find and execute the file as "perl mytest.pl", passing a small number of environment variables with information on, for example, a temporary directory it can use if necessary. If any extra command line args had been given they would also be transmitted to the test for it to use in any way it likes. As long as mytest.pl reports test progress using TAP, all is well. At this point, any number of other scripts can be added, including organizing them in whatever structure makes sense for you.
However, the test suite should preferably as soon as possible be given a config file in the root of the suite.
A test suite is configured by putting the file config.testontap in the root of the suite.
The standard module used for reading this file is Config::Std, please consult this module for general information on the format. In particular, some keys allows multi-line and/or multi-part values.
Also, in several places this file can contain queries (e.g. to select tests) using a query format as described in Grep::Query. Notable here is that queries operate on paths inside the suite, so those are defined as simply the relative path of the file from the root of the suite, and always using '/' as a path delimiter on all platforms. Directories will be suffixed with a '/'.
Generally, this file should at a minimum always contain a line of "id = <UUID>". In fact, if no config file is present, a warning will be displayed to the effect that a newly generated id will be used for each run. See below.
Empty lines are ignored, and comments can be made by starting the line with '#'.
The following keys/sections are valid:
This is required and must be set to an UUID.
It will be used to mark results with and thus can be used to distinguish persisted results from different suites. Hence, you should obviously not reuse an id already used for another suite! It is however expected that something processing such results are aware that a suite most likely will change during its lifetime.
Unless you have any other tool to generate an UUID with, a tip is to run the suite once without a config file and then cut and paste the generated id.
Example:
id = 61d4c8c3-c827-11e6-a3f8-bea3de08f558
This is optional. The value is a Grep::Query expression, and can be multi-line/part for readability.
If not present, it effectively defaults to include all files, assuming they match given extensions in the "exec mapping" (see below).
This is useful for excluding parts of the suite tree, e.g. parts which are merely data to be used by the tests and that may contain files with extensions that would otherwise be mistaken for active tests.
As the suite is scanned for tests, paths will be tested against the query, and if not selected, will be pruned completely from the set of possible tests. In particular, not selecting a directory will prune the entire subtree.
include = not eq(unfinished_test.pl) and not regexp{^(.*/)?_data/$}
This would avoid the 'unfinished_test.pl' file and contents of any '_data' directories at any level of the suite.
This is optional, but necessary if parallelization using the '--jobs' flag is to be effective (a warning will be generated if the user requests a --jobs value greater than one, but no parallelizable rule is present). The value is a Grep::Query expression, and can be multi-line/part for readability.
If not given, it effectively defaults to forcing all tests to run serially, on the theory that a test not written with the understanding that it could be run in parallel with other tests might end up producing weird results. Hence, it forces the test suite developer to think about this and knowingly turn on parallelizability.
As tests are evaluated as being ready to run, the names will also be matched against the query to evaluate for being allowed to run in parallel. Note that at this point, it will always be files, never directories.
Also note that tests may also be subject to dependency rules (see below), which may or may not impact the ability for them to run in parallel with anything else.
Examples:
parallelizable = regexp(.*)
In one stroke, this will allow all tests to be parallelizable. Obviously, using 'not' in front, will create a query that ends up acting as the default, i.e. tests are never parallelizable (but it will shut up the --jobs warning).
parallelizable = not eq(some/path/lonely_test.pl) or not eq(another/path/singleton.pl)
E.g. all tests are possible to run in parallel except two special ones.
This section is optional, but recommended. It details how files with particular file extensions should be executed. Note that there is also possible to pass a file with this section manually using the '--execmap' option, in order to override or complement any defined execmap.
The section should contain key/values where the key lists one or more extensions and the value details the command to use for those extensions. The value may be in multi-line/part format, which is required if the command contains whitespace in the individual parts. A plain string will be split on space in order to separate the command into parts.
[EXECMAP] pl t = perl jar = java = -jar # in this case, could also have been written simpler as 'jar = java -jar'
This means that any files found with the extensions 'pl' or 't' will be run by 'perl <file>' and that any 'jar' files will be run by 'java -jar <file>'. Any other files will be ignored.
If no "EXECMAP" section is present, a default one will be used, equivalent to this for this release:
[EXECMAP] pl t = perl jar = java = -jar py = python groovy gsh gvy gy = groovy # only on windows bat cmd = cmd.exe = /c # only on unix/linux sh = /bin/sh
While useful in the early stages for quickly getting started, good form is to ensure the execmap is explicit, avoiding surprises if other extensions are added to the default, and also to use the "include" key to ensure only expected files are selected as tests regardless.
Sections like these are optional. There can be many with different names (the names have no internal meaning beyond helping when reporting problems with dependencies).
Each section describes dependencies between one or more tests on one or more other tests; it will ensure that any dependencies are executed first. This can be utilized for example by having a test that ensures a database is set up in a specific manner before proceeding.
Note that any given test can be selected by multiple dependency sections, in either match/dependson, and this will ultimately create a dependency graph. However, it must result in a "DAG", no cycles are allowed.
A dependency section has two required key/value pairs:
This is required. The value is a Grep::Query expression, and can be multi-line/part for readability.
It must select one or more tests that will be dependent on tests selected by the "dependson" key.
It must select one or more tests that must be run before the tests selected by the "match" key.
[DEPENDENCY All depends on init] match = not eq(init.pl) dependson = eq(init.pl) [DEPENDENCY File handling] match = regexp(^filehandl/) dependson = eq(basic_file_handling_setup.pl) [DEPENDENCY Picture processing] match = regexp(^picproc/) dependson = eq(basic_picture_processing_setup.pl)
The three separate rules will together create a dependency tree between tests. The first is basically simple: all tests are dependent on the 'init.pl' test (except itself, obviously). The effect is that init.pl will run to completion first before any other test may start. This will be true regardless of if parallelization is in effect or not.
Further, the next two rules ensure that the two different 'basic_...' tests run after that. They are independent, so they can run in parallel. As they complete, all tests under 'filehandl/' and 'picproc/' are released for running, and again, all of them are independent and can run in parallel.
Again, beyond obeying the dependencies, exactly which test(s) will run in what order are ultimately dependent on many other factors such as the 'parallelizability' of a given test, how many job slots are available, exactly how the TAP harness implementation handles parallelization and how capable it is etc.
Most test suites will need some kind of outside input when running. There are two main sources:
Examples of this is simply to allow the tests to be flexible. The most obvious example is that they need information on where/how to access the SUT or similar. While it could be hardcoded into the suite in some manner, it may make sense to make this information into something that is passed by the user when starting the suite, thus allowing the user to point to several different SUT's using the same test suite and evaluate the behavior of each.
To accomplish this, the recommended way is to simply pass this information along on the command line, for example:
testontap --jobs 3 some/path/to/fancysuite SUT=xyz VERSION=BETA2 "my sample test run"
The parameters after the suite path is not processed by testontap in any way but will all be passed as command line arguments to each test.
As the tests can interpret the args on the command line in any manner they choose, for more advanced requirements consider using a suitable file format to pass in your humongous and/or complex set of params and just pass the file name.
As the driver controls the tests, it also controls various disk locations which the tests are expected to use for temporary needs as well as some data it wants to persist. By following these pointers, the driver may package the data cleanly and finally clean up after completion.
The driver communicates this type of information using environment variables (prefixed by 'TESTONTAP_'):
TESTONTAP_TMP_DIR
This variable holds an existing directory path, and is free for the tests to use for whatever reason. Note that it will be the same for all tests in the suite and empty when the first test runs, but will not be touched until all tests have finished - hence, it can be used to pass information between tests in the suite (however, consider parallelization issues if this is used).
When all tests are concluded, it will be removed; no data in it will be persisted.
TESTONTAP_PRIVATE_DIR
This variable holds an existing directory path, and is free for the tests to use for whatever reason similar to the above temporary directory. The difference is that files/directories in here will be persisted along with the general results.
When all tests are concluded, data in it will be persisted before it is removed.
The results can be persisted as either a plain directory tree or a zip archive. The naming will consist of the test suite name, the timestamp when the test was run and a 'run id' (an UUID, generated for each test run). If it's an archive file it will be suffixed with '.zip'. See command line options and arguments for details on how to control this using '--archive' and '--savedirectory'.
The saved results are actually a part of the 'work directory', the place where all files are managed until they're saved. This includes, among other things, the temporary directory used for "TESTONTAP_TMP_DIR" and may therefore sometimes be useful in debugging. It is normally a temporary location and automatically cleaned away, but by using the '--workdirectory' option, it will remain after completion so you may examine it.
testontap some/path/to/fancysuite
This will only run the suite, but will not persist any results.
testontap --archive some/path/to/fancysuite
This will persist the test results as the file fancysuite.20161227T103707Z.6a00a78d-cc20-11e6-bceb-fc7c2814a673.zip in the current directory.
testontap --archive --savedirectory /tmp some/path/to/fancysuite
This will persist the test results as the file fancysuite.20161227T103707Z.6a00a78d-cc20-11e6-bceb-fc7c2814a673.zip in the /tmp directory.
testontap --savedirectory /tmp some/path/to/fancysuite
This will persist the test results as the directory fancysuite.20161227T103707Z.6a00a78d-cc20-11e6-bceb-fc7c2814a673 in the /tmp directory.
Note that for the zip file, it will internally contain a root directory with the same name.
Assuming a test suite looking like below, and also assume this suite will generate a private persistable 'ioperf.xml':
fancysuite/ -- config.testontap -- init.pl -- basic_file_handling_setup.pl -- basic_picture_processing_setup.pl -- filehandl/ -- -- fh1.pl -- -- fh2.pl -- picproc/ -- -- pp1.pl -- -- pp2.pl
Running the suite using:
testontap --savedirectory /tmp fancysuite
The test suite result tree in /tmp will look like this:
fancysuite.20161227T103707Z.6a00a78d-cc20-11e6-bceb-fc7c2814a673 -- env.json -- meta.json -- summary.json -- private/ -- -- ioperf.xml -- result/ -- -- init.pl.json -- -- basic_file_handling_setup.pl.json -- -- basic_picture_processing_setup.pl.json -- -- filehandl/ -- -- -- fh1.pl.json -- -- -- fh2.pl.json -- -- picproc/ -- -- -- pp1.pl.json -- -- -- pp2.pl.json -- tap/ -- -- init.pl.tap -- -- basic_file_handling_setup.pl.tap -- -- basic_picture_processing_setup.pl.tap -- -- filehandl/ -- -- -- fh1.pl.tap -- -- -- fh2.pl.tap -- -- picproc/ -- -- -- pp1.pl.tap -- -- -- pp2.pl.tap
General note is that 'json' files are in UTF-8.
The described format version here is '0.1', and is indicated in the meta-data below (as format => major, minor).
env.json
This contains a dump of the enviroment when the suite was run, as key/value pairs.
meta.json
This contains general information about the test run.
argv (list of strings)
The arguments given on the command line and passed to the tests.
begin (string)
The moment when the test run started, as a UTC timestamp in ISO8601 format.
defines (key/value pairs)
Any defines given when invoking the testontap command.
elapsed (key/value pairs)
CPU and real clock values for the run:
cpu (number)
Used total CPU time.
real (number)
Real elapsed time "wallclock seconds".
str (string)
Elapsed time in display form.
end (string)
The moment when the test run ended, as a UTC timestamp in ISO8601 format.
format (key/value pairs)
The format version of this result tree. If any changes are made, this will be updated and changes documented.
major (number)
Incremented for incompatible changes, e.g. removal of a file, changed semantics etc.
minor (number)
Incremented for backwards compatible changes, e.g. the addition of a new file, key/value pair etc.
host (string)
The FQDN for the host where the test was run.
jobs (number)
The job slots used when the test was run.
platform (string)
The platform of the host where the test was run.
runid (string)
The UUID for the test run (generated every time).
suiteid (string)
The UUID for the test suite (as defined in the suite configuration).
suitename (string)
The name of the suite, defined as the name of the directory.
uname (list of strings)
Data from the uname() function.
user (string)
The name of the user that ran the test.
summary.json
This holds a summary of the entire test suite, with information coming from the aggregator (see TAP::Parser::Aggregator).
all_passed (0 or 1)
True (1) if all tests passed.
failed (list)
Tests that failed.
parse_errors (list)
Tests generating invalid TAP.
passed (list)
Tests that passed.
planned (list)
Tests planned to run.
skipped (list)
Tests skipping microtests.
status (PASS or FAIL)
Overall status.
todo (list)
Tests having TODO's.
todo_passed (list)
Tests with TODO's that unexpectedly passed.
private/
Any files in this directory are written by the test suite itself and is defined by that.
result/
The structure below this directory mimics the structure in the suite. Every test is represented by a file with the extension '.json', i.e. a JSON file of key/value pairs of data about the microtests in the test.
actual_failed (list)
The microtests that failed regardless of if they we're TODO's.
actual_passed (list)
The microtests that passed regardless of if they we're TODO's.
end_time (string)
The moment when the test ended, as a UTC timestamp in ISO8601 format (in high resolution time, if available).
exit (number)
The exit code from the test.
The microtests that failed, but if it was a TODO, it isn't counted.
has_problems (0 or 1)
True (1) if there are any problems with this test.
is_good_plan (0 or 1)
True (1) if the plan corresponded to the number of microtests.
Lists of microtests that generated erroneous TAP.
The microtests that passed, even if it is a failing TODO.
plan (list)
The test plan.
skip_all (0 or string)
True (reason string) if skipping everything.
Individual microtests skipped.
start_time (string)
The moment when the test started, as a UTC timestamp in ISO8601 format (in high resolution time, if available).
tests_planned (number)
The number of microtests planned.
tests_run (number)
The number of microtests run.
Microtests with TODO directives.
The microtests with TODO directives passing.
version (number)
The TAP version detected.
tap/
The structure below this directory mimics the structure in the suite. Every test is represented by a file with the extension '.tap', and is the actual output given by the test (which ought to be TAP, but in case of a broken test or an executable mistaken for a test, may show non-existent or broken TAP, and if so, should show up with parse errors in the summary).
Kenneth Olwing, <knth at cpan.org>
<knth at cpan.org>
Please report any bugs or feature requests to bug-app-testontap at rt.cpan.org, or through the web interface at http://rt.cpan.org/NoAuth/ReportBug.html?Queue=App-TestOnTap. I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.
bug-app-testontap at rt.cpan.org
You can find documentation for this module with the perldoc command.
perldoc App::TestOnTap
You can also look for information at:
RT: CPAN's request tracker (report bugs here)
http://rt.cpan.org/NoAuth/Bugs.html?Dist=App-TestOnTap
AnnoCPAN: Annotated CPAN documentation
http://annocpan.org/dist/App-TestOnTap
CPAN Ratings
http://cpanratings.perl.org/d/App-TestOnTap
Search CPAN
http://metacpan.org/dist/App-TestOnTap/
I thank my family for putting up with me!
https://github.com/kenneth-olwing/App-TestOnTap.
Copyright 2017 Kenneth Olwing.
This program is free software; you can redistribute it and/or modify it under the terms of the the Artistic License (2.0). You may obtain a copy of the full license at:
http://www.perlfoundation.org/artistic_license_2_0
Any use, modification, and distribution of the Standard or Modified Versions is governed by this Artistic License. By using, modifying or distributing the Package, you accept this license. Do not use, modify, or distribute the Package, if you do not accept this license.
If your Modified Version has been derived from a Modified Version made by someone other than you, you are nevertheless required to ensure that your Modified Version complies with the requirements of this license.
This license does not grant you the right to use any trademark, service mark, tradename, or logo of the Copyright Holder.
This license includes the non-exclusive, worldwide, free-of-charge patent license to make, have made, use, offer to sell, sell, import and otherwise transfer the Package with respect to any patent claims licensable by the Copyright Holder that are necessarily infringed by the Package. If you institute patent litigation (including a cross-claim or counterclaim) against any party alleging that the Package constitutes direct or contributory patent infringement, then this Artistic License to you shall terminate on the date that such litigation is filed.
Disclaimer of Warranty: THE PACKAGE IS PROVIDED BY THE COPYRIGHT HOLDER AND CONTRIBUTORS "AS IS' AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES. THE IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT ARE DISCLAIMED TO THE EXTENT PERMITTED BY YOUR LOCAL LAW. UNLESS REQUIRED BY LAW, NO COPYRIGHT HOLDER OR CONTRIBUTOR WILL BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING IN ANY WAY OUT OF THE USE OF THE PACKAGE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
To install App::TestOnTap, copy and paste the appropriate command in to your terminal.
cpanm
cpanm App::TestOnTap
CPAN shell
perl -MCPAN -e shell install App::TestOnTap
For more information on module installation, please visit the detailed CPAN module installation guide.