Code Coverage

This is the state of the test cases coverage as of 2016-06-14.

Total number of branches:    3877
Number of covered branches:  3027
Percent branch coverage:     78.1%
Total lines of code:        25219
Covered lines of code:      19735
Percent LOC coverage:        78.3%

Barcode Type	Total Branches	Branches Covered	Percent
auspost	32	32	100%
azteccode	135	128	95%
azteccodecompact	2	2	100%
aztecrune	2	2	100%
bc412	23	16	70%
channelcode	25	19	76%
codablockf	117	81	69%
code11	30	16	53%
code128	89	72	81%
code16k	142	72	51%
code2of5	23	22	96%
code32	14	10	71%
code39	24	24	100%
code39ext	8	7	88%
code49	76	48	63%
code93	18	17	94%
code93ext	6	6	100%
codeone	260	135	52%
coop2of5	2	2	100%
daft	10	8	80%
databarexpanded	209	149	71%
databarexpandedcomposite	15	13	87%
databarexpandedstacked	2	2	100%
databarexpandedstackedcomposite	13	12	92%
databarlimited	56	51	91%
databarlimitedcomposite	5	3	60%
databaromni	98	93	95%
databaromnicomposite	14	12	86%
databarstacked	10	6	60%
databarstackedcomposite	12	9	75%
databarstackedomni	10	6	60%
databarstackedomnicomposite	12	9	75%
databartruncated	10	6	60%
databartruncatedcomposite	14	12	86%
datalogic2of5	2	2	100%
datamatrix	315	265	84%
datamatrixrectangular	2	1	50%
ean13	42	42	100%
ean13composite	6	5	83%
ean14	14	9	64%
ean2	16	16	100%
ean5	19	19	100%
ean8	40	40	100%
ean8composite	6	5	83%
flattermarken	10	8	80%
gs1-128	24	23	96%
gs1-128composite	13	11	85%
gs1-cc	181	103	57%
gs1datamatrix	21	19	90%
gs1datamatrixrectangular	21	19	90%
gs1qrcode	21	19	90%
hanxin	109	98	90%
hibcazteccode	9	4	44%
hibccodablockf	9	4	44%
hibccode128	10	5	50%
hibccode39	10	5	50%
hibcdatamatrix	9	4	44%
hibcdatamatrixrectangular	9	4	44%
hibcmicropdf417	9	4	44%
hibcpdf417	9	4	44%
hibcqrcode	9	4	44%
iata2of5	2	2	100%
identcode	10	7	70%
industrial2of5	2	2	100%
interleaved2of5	17	15	88%
isbn	57	53	93%
ismn	50	29	58%
issn	33	19	58%
itf14	13	10	77%
japanpost	18	17	94%
kix	13	12	92%
leitcode	10	7	70%
matrix2of5	2	2	100%
maxicode	85	58	68%
micropdf417	127	96	76%
microqrcode	2	1	50%
msi	30	28	93%
onecode	47	43	91%
pdf417	130	118	91%
pdf417compact	2	2	100%
pharmacode	12	9	75%
pharmacode2	11	8	73%
planet	19	14	74%
plessey	20	15	75%
posicode	64	40	63%
postnet	19	14	74%
pzn	12	5	42%
qrcode	285	256	90%
rationalizedCodabar	29	19	66%
raw	7	6	86%
royalmail	19	18	95%
sscc18	14	10	71%
symbol	7	7	100%
telepen	16	13	81%
telepennumeric	2	2	100%
upca	56	56	100%
upcacomposite	6	5	83%
upce	58	58	100%
upcecomposite	9	8	89%

If there is a barcode encoder that has low test coverage that you would like to adopt, please help out.

To run the tests, you will first need to follow the instructions in Compiling bwipjs. Make sure to install ghostscript and create the gs wrapper script, if necessary.

Next we need to create test cases. Take a look at the bwip-js/runtests script to see the existing test cases. The format reflects what you would enter in the bwipjs demo:

run <encoder> "<barcode text>" "<options>"

Then click on the encoder link in the table above. That takes you to the code coverage map for the encoder. Black text has been covered. Red text needs a test case.

Low priority are the branches that contain a call to raiseerror (although test cases for them will be happily accepted).

Highest priority are forall loops that have either of the following structures:

// With callback
$forall(<expression>, function() {
	<lots of red code here>
});

// Without callback
$forall(<expression>);

Those loops implement a slow-path as the trace compiler was not able to properly analyze the type of object in the forall-expression and had to let the runtime library determine the correct postscript semantics.

The goal is to have test coverage for all of these worst-case loops. The cross-compiler can then be enhanced with runtime feedback that allows it to emit optimized code.

Next highest priority are loop bodies with any red code. Loops can be tricky in postscript and may manipulate the operand stack in unexpected ways. Each one needs testing.

Trying to determine a test case that will exercise a branch of code is a bit of an art. Postscript is not an intuitive language and has a fairly steep learning curve. The code generated by the cross-compiler, in my opinion, is actually easier to read than the original postscript (Terry will likely disagree!).

Before running your test cases, you need to build bwipjs with code coverage enabled. Simply:

$ ./psc --with-coverage

The new bwipp.js will contain embedded logic to record branch coverage at runtime. This code is not as performant as the standard bwipp.js code so remember to re-compile your code without coverage after you are done:

$ ./psc  # do this only after you are done with code coverage testing

Now, add your test cases to runtests under the existing test cases for the encoder. Run the tests and generate a new coverage map:

$ ./runtests <encoder>
$ ./uncover

Where <encoder> is the name of the encoder you added test cases for.

uncover will update covermap.html in your local directory. Go to your browser and use a file:/// URL to navigate to the page. For example, if you installed bwipjs under C:\, the URL will be:

file:///c:/bwip-js/covermap.html#encoder

Where encoder is the name of the encoder. Verify that the test cases covered the code branches you were targetting.

Finally, use the demo to verify that your test cases are either generating a barcode or the expected error message.

Please submit your tests cases as a new issue on the project's github pages. I will not take pull requests for the runtests file as I often modify it and only push changes when releasing.

After you submit your tests cases, a 2D bar code scanner is used to verify the barcodes are readable and contain the correct data. If all goes ok, your test cases (and my warm thanks!) will be part of the next release.