Python-based TinyG tester for v8 and g2core code bases. The tester runs JSON test files and reports results. This project is maintained in the WingIDE Python but can use any Python environment.
Please note: tg_pytest is only intended to get the easy tests done; It is not intended to be a full-featured test framework. It is not suited to running continuous tests, i.e. running a Gcode file and looking for the ongoing results. It's a batch tester, not streaming.
-
The tester opens the test-master.cfg file and runs each uncommented JSON file in sequence.
-
Each JSON file contains one or more tests structured as standalone JSON objects.
-
Each test is run by sending all the strings in the
send
array, then waiting for all the JSON responses to be returned from the board. Responses may consist ofr
,sr
ander
JSON objects. Responses are decoded and put in a list of decoded JSON objects. Sent strings are displayed as--->
, responses as<---
-
After a 1 second timeout (settable in tg_utils.py) the tester stops listening and analyzes the response list. A separate analyzer is called for
r
,sr
ander
responses. Results are sent to the screen (and in the future to an optional file). -
Note:
sr
responses are rolled up into a "synthetic" status report, which is the union of all elements received by all status reports. The last instance of an element is what sticks (although it still can be useful set status reports to verbose). -
See
/data/000-init/example-init.json
for examples of items in this readme.
- Make sure you have a TinyG v8 or g2core powered and plugged in and it's not already connected to some other USB host (like Coolterm or g2core-api)
- Edit the
/data/.../test-master.cfg
file for the tests you want to run - Make sure each JSON test file is correct (see below)
- Run tg_pytest from your Python environment
- Note: For v8, the board must be in strict JSON mode or the Python JSON parser will fail
- Note: For g2core, the board must be in {ej:1} or {ej:2} or the Python JSON parser will fail
- Opens a g2core or TinyG port or fail trying (only one device may be connected)
- Opens test-master.cfg file
- For each test in the master file:
- Opens the JSON file for that test
- Creates an output file suffixed by aN ISO8601 date/time string (future)
- Iterates through the input file. For each test in file:
- Send the
before
strings to the board - Send the
send
string(s) to the board and collect all responses - Analyze responses according to
t
test data provided (see below) - Send the
after
strings - Note that only
send
strings are analyzed.before
andafter
are not
- Send the
- A JSON test file contains one or more tests
- Each test is defined by an independent JSON object
- JSON objects must be separated by at least one comment line
- Comment lines start with
#
- Open curlies
{
are not allowed in comments - Adding
SKIP
to a comment line will skip the next test - Adding
EOF
to a comment line will end the file - Blank lines are permitted but do not act as separators between JSON objects
- A test
t
runs all thesend
lines, collects all the response linesr{}
,sr{}
,er{}
then analyzes according to the analyzers: JSON tagsr
,sr
,er
- Supported JSON elements are below, and are not order dependent. Arrays are order dependent.
- Each JSON test must be a parseable JSON object or the test will fail to execute. If in doubt, lint it.
- It's useful to use a linting editor like Atom to edit JSON files (use json-linter package)
- It can also be useful to paste a JSON object into jsonlint.com for checking
- JSON syntax errors in either the test JSON or the response JSON (for the board) will cause the test to fail with a message including the offending JSON.
JSON Elements
-
t
is the test data object, consisting of: (all are optional exceptsend
)label
will be displayed when the test is runsend
array of one or more strings to send for the test (MANDATORY)delay
delay in seconds between sends. Values < 1.0 are OKprecision
is the precision to match for numerical valuesfail
can behard
orsoft
. Hard (default) quits the test run if failurefail_returned_er
fail if aner
exception report is in the responsefail_missing_sr
fail if ansr
element specified insr
object is missing from responsebefore
array of strings to send before the test - will not be analyzedafter
array of strings to send after the test - will not be analyzedsetup
iftrue
, prevent before's and after's from executing in a given test
-
r
analyzer contains the elements to check in allr{}
responses:- If
r
is present, test all listed keys for a match. One of:- Exact match
- Match to precision set by
precision
- Wildcard match
"*"
will match any value (quotes are required)
- Use
status
to match the status in the footer - Use
count
to match the count in the footer (3rd element) - Nested keys can also be tested, example:
"send":["{sr:n}"]},
can be tested for:"r":{"status":0, "sr":{"posx":0.0"}}
- If
-
sr
analyzer contains the elements to check in the status reports:- If
sr
is present, test all keys for exact match or precision match, e.g.stat:3
- If
-
er
analyzer contains the elements to check in any exception reports- Any ERs thrown will be displayed
- ERs will be displayed by default unless disabled by
display:false
- No elements are actually matched
Note that strings in embedded JSON do not need to be escaped, as TinyG will always accept JSON in relaxed mode, regardless of whether it's set to relaxed or strict JSON mode. (The board must be in strict mode so all responses will be strict, of course). The shortcuts also work, like t
for true
or n
for null
. The following example still produces valid JSON but is easier to edit and maintain:
{
"t":{"label":"Read axis configuration",
"send":["{\"x\":null}"]} ...can be sent as:
}
{
"t":{"label":"Read axis configuration",
"send":["{x:n}"]}
}
Another trick is to "comment out" a line in a JSON file just invalidate the tag by putting XXX_
or some other characters (but not #) in front of it. The tag/value will be ignored by the tester.
Additional JSON objects can be provided in a file that will run before and after tests. The JSON format is the same as the t
object, but only the listed tags are recognized. before_each
will run prior to local before
, and after_each
will run following local after
in a given test object.
-
before_all
will run once before all tests in a file:label
display the label in the outputbefore
array of strings to send at start of filedelay
optional delay in seconds
-
after_all
will run once after all tests in a file: -
label
display the label in the output -
after
array of strings to send after file is complete -
delay
optional delay in seconds -
before_each
will be run before each test object in a file:label
is optional, but is not displayedbefore
array of strings to send before each test before localbefore
delay
optional delay in seconds
-
after_each
will be run after each test object in a file:label
is optional, but is not displayedafter
array of strings to send after each test after localafter
delay
optional delay in seconds
-
setup:true
-
can be included in a test to prevent
before_each
,after_each
and localbefore
andafter
from being run. This supports building setup "tests" that run before the actual tests in the file. (Note: this tag may evolve to something with more fine-grained control)
It's possible to add a delay between strings in a send array. For example:
"send":["m3 m7 m8", "G1 F500 X30", "del 0.25", "!"]
...inserts a 250 ms delay before the bang. The space before the value is required.
A defaults
JSON object can be included in a test file. Defaults are read and applied in the location they are found in the file, so it's good to list these first, or sometimes after the before_all. The default settings will be applied to all subsequent tests. Local, per-test settings for the same variables will override the defaults. Values supported include:
fail
set assoft
orhard
. Hard fail will stop tests immediately on a failure, Defaults tohard
fail_returned_er
fail if aner
exception report is in the responsefail_missing_sr
fail if ansr
element specified insr
object is missing from responsedelay
delay in seconds between sends. Values < 1 are OKprecision
for matching numeric values can be set in the defaults
- Tests are run in the order listed in
/data/test-master.cfg
- Tests can be commented in and out using
#
in the first char. - A test run should start with a setup file such as
/data/000-setup/setup-centered-001.json
- Tests can be designed to run on a naked board (board and motors) or on a machine. If the test is run on the machine some guidance should be provided for the initial conditions. It's common practice to run a test from the middle of travel - i.e. 1/2 way between limits on X, Y and Z. See arc tests for examples.
- Some limited flow control is provided for TX serial buffer management when transmitting the
before
,send
, orafter
string arrays. The serial port is set to RTS/CTS, but we have noticed that Pyserial does not reliably use this flow control across all platforms. So it's best if the total byte count of strings sent to the board be limited to 254 characters (v8 streaming mode), or 10 lines (v8 line mode), or 3000 characters (g2). In most tests this should not be an issue. - The analyzers only support exact-match checking of results. This should be enough for now.
- As of 3/1/16 the USB port finder now works on OSX and Windows. It should also work on Linux but has not been tested.
- Added before, after and SKIP behaviors 3/8/16
- Added file include function 3/26/16
- document file include (or perhaps import)
- make include work only as from first non-space character
- make JSON file importer smart enough to find break between JSON objects