The Plax Manual

To vouch this, is no proof, without more wider and more overt test.

Installation

To build Plax, you need Go installed.

Once you have a working go environment, you can install plax using the following commands from the root of this repository.

go get github.com/Comcast/plax/...

Check that you can execute plax:

Using Plax

Basic use

plax -h

plax -h
Usage of plax:
  -I value
    	YAML include directories
  -channel-types
    	List known channel types and then exit
  -dir string
    	Directory containing test specs
  -error-exit-code
    	Return non-zero on any test failure
  -json
    	Emit docs suitable for indexing
  -labels string
    	Optional list of required test labels
  -list
    	Show report of known tests; don't run anything.  Assumes -dir.
  -log string
    	log level (info, debug, none) (default "info")
  -p value
    	Parameter values: PARAM=VALUE
  -priority int
    	Optional lowest priority (where larger numbers mean lower priority!); negative means all (default -1)
  -redact
    	Use redaction gear (default true)
  -retry string
    	Specify retries: number or {"N":N,"Delay":"1s","DelayFactor":1.5}
  -seed int
    	Seed for random number generator
  -test string
    	Filename for test specification
  -check-redact string
    	input string to use for -check-redact-regexp
  -check-redact-regexp string
    	regular expression to use for checking redactions
  -test-suite string
    	Name for JUnit test suite (default "NA")
  -v	Verbosity (default true)
  -version
    	Print version and then exit

The plax command can run a single test or a set of test.

To run a single test, use -test:

plax -test demos/simple.yaml -log debug

To run tests in a directory, use -dir DIRNAME. When using -dir, you can also specify a comma-separated list of required labels with -labels. For example, to run tests that are labeled foo or bar:

plax -dir demos -labels selftest

You can also specific a minimum priority via -priority. For example:

plax -dir demos -priority 3 -labels selftest

will run all selftest tests in the demos directory that that a priority less than or equal to 3.

You can pass bindings in the command line using -p. You can specify multiple -p values:

plax -test foo.yaml -p '?!WANT=tacos' -p '?!N=3'

Using `plaxrun`

For more sophisticated Plax test execution, see the plaxrun manual, which documents using plaxrun to run lots of Plax tests under various configurations.

Writing Tests

You write a test specification in YAML. This section describes the structure of a test specification. Also see these examples. basic.yaml is a good, small example of a test specification.

Channel types

A Plax test does I/O using "channels". Currently Plax supports the following channel types:

mqtt: An MQTT client
kds: A primitive KDS consumer
sqs: A basic SQS consumer and publisher
httpclient: An HTTP client
httpserver: An HTTP server
cmd: Shell I/O
mock: an echoing channel for testing
cwl: A Cloudwatch Log publisher and consumer

As the needs arise, we can add channel types like:

KDS publisher
Kafka consumer and publisher

and so on.

The plax executable supports -channel-types to list the known channel types and then exit.

Including YAML in other YAML

Plax supports including some YAML in other YAML.

Plax looks for:

#include<FILENAME>
$include<FILENAME>
include: FILENAME
includes: [FILENAME-1, FILENAME-N]

in your YAML. When Plax encounters one of these directives, Plax attempts to read FILENAME from directories specified on the command line via -I DIR. You can specify multiple -I DIR arguments. The test spec's current directory is added to the end of the list of directories to search. Then YAML that's read is substituted for the include directive.

For include: FILENAME or includes: FILENAME should be a YAML representation of a map. That map is added to the map that contained the include: FILENAME property.

$include<FILENAME>, which must be a value in an array, results in a splice into that array by the array represented by the YAML in FILENAME. Unlike cpp, Plax looks for FILENAME relative to the test's directory.

#include<FILENAME> is replaced by that value with the thing represented by FILENAME in YAML. Unlike cpp, Plax looks for FILENAME relative to the test's directory.

The utility command yamlincl performs just this processing. Example:

cat demos/include.yaml | yamlincl -I demos

Name

The optional name field is used for giving a concise identifier for a test. The value isn't actually used for anything at the moment.

name: discovery-1

Labels

The optional label field is used to list general attributes of or tags for the test. For example, what components are tested or what type of test it is. A label can be any string, but we shouldn't go crazy here. The plax tool's -label option can run only tests that that all of the given labels (separated by commas). For example plax -dir tests -labels integration,happy-path would run all the tests in the directory tests that have labels integration and happy-path.

Example test labels:

labels:
  - happy-path
  - integration
  - authentication

Priority

The optional priority field assigns a priority to the test. Priority is used to select which tests to run. Priority can be passed into plax using the -priority option. For example, when a priority of 2 is passed into plax, it will run all level 1 and level 2 priority tests, but not level 3.

priority: 1

Documentation strings

The optional doc attribute can provide documentation as a string. Note that YAML supports multi-line strings, and your doc value should probably be one of those.

We might add a links attribute that could specify a list of URLs of interest.

Negative

The optional negative field means a failure is interpreted as a success, and a failure is interpreted as a success. Errors (as opposed to failures) are not affected.

Example:

negative: true

Retries

The optional retries field specifies a retry policy:

n: The maximum number of retries
delay: The initial delay (in Go syntax)
delayfactor: A multiplier to applied to a delay to give the next delay.

The default behavior is no retry.

Example:

retries:
  n: 3
  delay: 1s
  delayfactor: 2

Bindings

Bindings allow a test to have values that change at runtime. For example, you could have a binding for a certificate filename that would allow you to run the same test with different filenames.

In an expression, bindings substitution takes two forms: structured and textual.

When processing a string, each occurrence of {B}, where B is a bound variable, is literally replaced by that variable's binding. For example, the string "I like {?x}." with with bindings {"?x":"queso"} results in "I like queso."

When processing structured data (which can be obtained implicitly from a string that's legal JSON), bindings substitution is itself structured. Only bindings starting with ? are considered, and only exact bindings are replaced. For example, the object {"need":"?x"} with bindings {"?x":"chips"} becomes {"need":"chips"}.

Note the difference between string-based bindings substitution and structured bindings substitution. The former results in a string value while the latter results in a value with the type of whatever the bindings value has. Plax will warn if you do bindings substitution in a string context where the binding value isn't itself a string.

All of this substitution is called recursively until a fixed point is reached, so you can go crazy with self-referencing substitutions. See demos/recursive-subst.yaml for a mild example.

If you've got a binding for a variable that you want to remove for subsequent steps, you can use a run step to remove the binding manually (say with delete(test.Bindings["?x"])). See this, this, and that example regarding unintentional bindings substitutions. In a recv step (see below), you can also specify clearbindings: true to ignore any existing bindings that do not start with ?!.

See the end of the next section regarding the order of operations.

To provide a binding at runtime, use the -p flag:

plax -test tests/this.yaml -p '?!CERT=that.pem' -p '?!KEY=key.pem'

These two bindings, for ?!CERT and ?!KEY, start with ?! to ensure that those bindings are not cleared when clearbindings is specified in a recv step. This recv behavior is described below.

When using -p to specify a binding, if the given value parses as JSON, then that parsed value is used as the binding value. This behavior is convenient when doing structured binding substitution.

String commands

Several substrings have special powers.

When Plax sees {@@FILENAME}, then Plax attempts to substitute the contents of the file with name FILENAME for that substring. When Plax sees a pattern or payload of the form @@FILENAME, the same thing happens. The file is read relative to the directory that contained the test specification.

When Plax sees {!!JAVASCRIPT!!}, then JAVASCRIPT is executed as Javascript, and the result replaces that substring. Bindings substitution applies. The value returned by this Javascript is substituted for string. When Plax sees a pattern or payload of the form !!JAVASCRIPT, then the same thing happens.

These string commands are processed in the order above: first @@ and then !!. (So a file's contents could start with !!, which would trigger Javascript execution.) Bindings are substituted after string processing. All of this substitution is called recursively until a fixed point is reached, so you can drive yourself crazy with self-referencing substitutions.

The documentation below mentions when a string has these special powers ("string commands"). Most strings have these powers.

Channels

A Plax test can work with multiple "channels" simultaneously. A channel is something can that do I/O, and an MQTT client is the classic example. We can also have channels for a KDS consumer, a KDS publisher, an HTTP client, an SQS consumer, and so on.

See Channel types for a summary of available types.

There is one primordial channel named mother. You can ask mother to make other channels for you by publishing (pub) a message to mother, who will always reply. Example of making a request and receiving the reply:

- pub:
    doc: Please make a mock channel.
    chan: mother
    payload:
      make:
        name: mock
        type: mock
- recv:
    doc: Check that our request succeeded.
    chan: mother
    pattern:
      succeed: true

The payload of the request should specify the name for the channel to be created, the type of the channel (e.g., mock, mqtt, cmd, etc), and an optional config for any channel options.

Note that a test might want to verify that a request to mother failed. For example, a request to mother to create an MQTT client with invalid credentials should fail. Authentication tests often have this form.

Javascript libraries

A test can specify libraries, which should be a list of filenames. Each file should contain Javascript. All of those files are loaded for each Javascript execution. Each file is read from the directory that contains the test spec.

Example:

libraries:
  - library.js
  - foo.js

That declaration will result in library.js and foo.js loaded before each run or guard.

Circuit breaker

A test specification can specify maxsteps, which defaults to 100. The test will fail if it takes more than this number of steps in total. This property is useful as a circuit breaker for an potential loop caused by a branch step.

Pattern matching

In a receive (recv) step (describe below), the given pattern is matched against incoming messages. This matching is Sheens message pattern matching. Here are some examples. You can maybe use go get github.com/Comcast/sheens/cmd/patmatch to experiment:

patmatch -p '{"want":"?x"}' -m '{"want":"queso","when":"now"}'

Specifications

The spec field is where most of the action will take place. Each phase in the phases consists of one or more steps. A step is a single operation. Currently the following steps are supported:

sub: Subscribe to a topic (filter).
1. chan: The name for the channel for this step.
2. pattern: The topic (or topic filter) for the subscription. If the value is a JSON string, the string is first parsed as JSON. Parameters and bindings substitution applies.
recv: Look for certain messages that have arrived.
1. chan: The name for the channel for this step.
2. topic: Optional: The expected message should arrive on this topic. Parameters and bindings substitution applies.
3. schema: An option URI for a JSON schema, which is then used to validate the in-coming message before any other processing.
4. serialization: How to deserialize in-coming payloads. Either string or JSON, and JSON is the default.
5. pattern: A pattern that the message must match. Parameters and bindings substitution applies. String commands are also available
  
  The pattern has this structure.
  
  All bindings for variables that start with ?* are removed before this pattern substitution.
  
  Alternately, give a regexp instead of a pattern.
6. regexp: A regular expression (instead of a pattern). A regular expression will probably be more convenient for receiving non-JSON input.
  
  A named group like (?P<foo>.*) match results in a new binding for a ?*foo. If the name starts with an uppercase rune as in Foo, then the variable will be ?Foo.
  
  See demos/regexp.yaml for an example.
7. clearbindings: If true, delete all test.Bindings for variables that do not start with ?!.
8. timeout: Optional timeout in Go syntax.
9. attempts: Optional number of (maximum) attempts when dequeuing a message for recv. If a topic is provided the number of attempts is for the given topic only
10. target: Target is an optional switch to specify what part of the incoming message is considered for matching.
  
  By default, only the payload is matched. If target is "message", then matching is performed against {"Topic":TOPIC,"Payload":PAYLOAD} which allows matching based on the topic of in-bound messages.
11. guard: Guard is optional Javascript that should return a boolean to indicate whether this recv has been satisfied.
  
  Parameter and bindings substitution applies, and string commands are also available
  
  The code is executed in a function body, and the code should 'return' a boolean or an expression of the form Failure(STRING). A boolean indicates whether the recv will succeed. A Failure will terminate the test immediately as failed.
  
  The following variables are bound in the global environment:
  1. bindingss: the set (array) of bindings returned by match().
  2. bindings (also bs): The first set of bindings returned by match(). Probably the only ones you care about.
  3. elapsed: the elapsed time in milliseconds since the last step.
  4. msg: the received message ({"topic":TOPIC,"payload":PAYLOAD}).
  5. test: The whole test object.
    
    In particular, your Javascript code can use test.State, which is a map from strings to anythings. You can use test.State to store data accessible by Javascript to be executed later. Also test.T, which is the time the previous step executed, is also available.
    
    test.Bindings is a map from pattern variables (e.g., ?foo) to values. The map is set after each successful recv pattern match to be the (first) set of bindings from that match. This map is used to replace any pattern variables in the payload of the next pub.
    
    With great power comes great responsibility.
  6. print: a function that prints its arguments to log output.
  7. redactRegexp: a function that compiles and adds a redaction regular expression from the given string.
    
    If the Regexp has no groups, all substrings that match the Regexp are redacted by replacing the substrings with <redacted>.
    
    For each named group with a name starting with redact, that group is redacted (for all matches).
    
    If there are groups but none has a name starting with redact, then the first matching (non-captured) group is redacted.
    
    You can use a plax command-line mode to check how a redaction regexp will work:
```
plax -check-redact-regexp 'love (really thin pancakes)' -check-redact "I love really thin pancakes."
I love <redacted>.
```
    See demos/redactions.yaml for some more examples.
  8. redactString: a function that compiles and adds a redaction pattern that matches the given string literally.
  9. fail(MSG): a function that immediately terminates a test as a failure (as opposed to being broken). The given MSG is the text of the failure. See demos/runfail.yaml for example use.
  10. Failure: a function that returns an object representing a failure with the argument as the failure message.
  11. match: Sheen's pattern matching function.
```
BINDINGSS = match(PATTERN,MSG,BINDINGS);
```
    1. PATTERN is a Javascript thing.
    2. MSG is a Javascript thing.
    3. BINDINGS is an Javascript object representing input bindings (often just {}).
    4. BINDINGSS is the set of set of bindings returned by match. (So that second S isn't really a typo.)
    If an error occurs, it's thrown.
    
    See demos/match.yaml for an example.
12. run: Executed Javascript just like guard except that the return value is ignored. Parameters and bindings substitution applies. String commands are also available
pub: Publish a message.
1. chan: The name for the channel for this step.
2. topic: Optional: The expected message should arrive on this topic. Parameters and bindings substitution applies.
3. serialization: How to serialize the payload. Either string or JSON, and JSON is the default.
4. payload: A pattern that the message must match. If the value is a JSON string, the string is first parsed as JSON. Parameters and bindings substitution applies. String commands are also available.
5. schema: An option URI for a JSON schema, which is then used to validate the out-going message.
6. run: Execute Javascript just like a recv's guard except that the return value is ignored. Parameters and bindings substitution applies. String commands are also available.
wait: Wait for the given number of milliseconds.
kill: Kill the step's channel ungracefully.
1. chan: The name for the channel for this step.
exec: Run a command. Structure is the same as for an initially command. See exec.yaml for a simple example. Parameters and bindings substitution applies.
reconnect: Attempt to reconnect the channel (even if still connected).
1. chan: The name for the channel for this step.
close: Close the channel. The channel is also removed.
1. chan: The name for the channel for this step.
run: Execute Javascript as in a recv's guard except that the return value is ignored. Parameters and bindings substitution applies.
branch: A fancy mechanism for (conditional) branching to another phase. Parameters and bindings substitution applies.

The value of a branch is Javascript code that should return the (name) of the next phase or the empty string (to continue with the current phase).

Example:
```
branch: |
  return 0 < test.State["need"] ? "here" : "there";
```
goto: Go to another phase.
doc: A documentation string for a step that's just that documentation string. Doesn't actually do anything.

Most steps have an optional chan field, which should name the channel for the step. A spec can declare a defaultchan that will be used for all steps. If your test has only one channel, then that channel is the default.

spec:
  defaultchan: cpe

You can also specify that a step is required to fail:

spec:
  phases:
    one:
      steps:
      - pub:
          topic: want
          payload: '"queso"'
        fails: true

Note that fail is specified at the same level as the type of step (pub, recv, etc.).

You can also specify that a step should be skipped by specifying skip: true in the step.

spec:
  phases:
    one:
      steps:
      - pub:
          topic: want
          payload: '"queso"'
        skip: true

Note that skip is specified at the same level as the type of step (pub, recv, etc.).

How you organize phases and steps is up to you.

You can specify your first phase using initialphase, which defaults to phase1:

spec:
  ...
  initialphase: boot

You can specify one or more "final" phases that are executed after the main test execution (starting with the initial phase) terminates regardless of any error encountered.

spec:
  ...
  finalphases:
    - cleanup1
	- cleanup1

These phases are executed in the given order regardless of any errors each might encounter. Note that a "final" phase can goto another phase. In that case, that target phase should (probably) not be included in the finalphases list.

See finally.yaml for a short example.

Output

After test execution, plax (or plaxrun) will by default output results in JUnit XML:

<testsuite tests="1" failures="0" errors="0">
  <testcase name="tests/discovery-1.yaml" status="executed" time="11"></testcase>
</testsuite>

For plax, use -test-suite NAME to specify the suite's name. For plaxrun a suite name will be generated.

For plax and plaxrun use -json to output a JSON representation of test result objects. This output includes the following for each test case:

[
  {
    "Type": "suite",
    "Time": "2020-12-02T21:33:09.0728586Z",
    "Tests": 1,
    "Passed": 1,
    "Failed": 0,
    "Errors": 0
  },
  {
    "Name": "/.../plax/demos/test-wait.yaml",
    "Status": "executed",
    "Skipped": null,
    "Error": null,
    "Failure": null,
    "Timestamp": "2020-12-02T21:33:09.077102Z",
    "Suite": "waitrun-0.0.1:wait-no-prompt:wait",
    "N": 0,
    "Type": "case",
    "State": {
      "then": "2020-12-02T21:33:09.0781375Z"
    }
  }
]

Logging

The -log command-line option accepts none (default), info, and debug. To provide some level of logging without printing some possibly sensitive information, info does not log payloads or bindings values. In contrast, -debug will by default log binding values and payloads. However, with -log debug, the flag -redact enables some log redactions:

Values with binding names that start with X_ (ignoring non-alphabetic prefix characters like ?) will be redacted from debug log output if -redact=true (default is true). If the redacted X_ values are needed for debugging, use -redact=false locally.
In a test, Javascript (usually executed via a run step) can add redactions using the functions redactRegexp and redactString. See documentation for redactRegexp above for more information.

See demos/redactions.yaml for an example of both techniques.

References

The plaxrun manual
Sheens message pattern matching, and some examples
Sheens could be used to implement more complex tests and simulations
TCL "expect"
YAML Wikipedia page and YAML multi-line strings in particular

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

manual.md

manual.md

The Plax Manual

Table of Contents

Installation

Using Plax

Basic use

Using `plaxrun`

Writing Tests

Channel types

Including YAML in other YAML

Name

Labels

Priority

Documentation strings

Negative

Retries

Bindings

String commands

Channels

Javascript libraries

Circuit breaker

Pattern matching

Specifications

Output

Logging

References

Files

manual.md

Latest commit

History

manual.md

File metadata and controls

The Plax Manual

Table of Contents

Installation

Using Plax

Basic use

Using plaxrun

Writing Tests

Channel types

Including YAML in other YAML

Name

Labels

Priority

Documentation strings

Negative

Retries

Bindings

String commands

Channels

Javascript libraries

Circuit breaker

Pattern matching

Specifications

Output

Logging

References

Using `plaxrun`