Improve warning messages in C++

[ischoegl]

Please fill in the issue number this pull request is fixing

Fixes #683

Changes proposed in this pull request

• introduce warn_user in C++ layer (in analogy to warn_deprecated)
• create uniform formatting for warnings
• add information on method raising the warning
• replace various occurrences of ad-hoc formatted writelog warnings with warn_user
• issue warn_user for Troe when T2 (i.e. T**) is zero

Notes

Some ck2cti/ck2yaml converters appear to create length-4 arrays for Troe parameters, which raises the warning that was suggested for #683 in the unit test suite. I.e.

test_surface_mech (cantera.test.test_convert.ck2ctiTest) ... UserWarning: 'Troe::init': Zero T2 value is suppressed.
ok
test_reaction_units (cantera.test.test_convert.ck2yamlTest) ... UserWarning: 'Troe::init': Zero T2 value is suppressed.
UserWarning: 'Troe::init': Zero T2 value is suppressed.
UserWarning: 'Troe::init': Zero T2 value is suppressed.
UserWarning: 'Troe::init': Zero T2 value is suppressed.
ok
test_surface_mech (cantera.test.test_convert.ck2yamlTest) ... UserWarning: 'Troe::init': Zero T2 value is suppressed.
ok

It's probably not worth fixing ck2cti, but it also affects ck2yaml. Edit: the YAML import issue is fixed.

Other thoughts

Long-term, it may make sense to use a logger that allows for a meaningful differentiation of levels, e.g. Boost.Log. The introduction of warn_user now will facilitate a central replacement in application.h/.cpp in the future.

[codecov]

Codecov Report

Merging #741 into master will decrease coverage by <.01%.
The diff coverage is 38.98%.

@@            Coverage Diff             @@
##           master     #741      +/-   ##
==========================================
- Coverage   70.84%   70.83%   -0.01%     
==========================================
  Files         372      372              
  Lines       43704    43720      +16     
==========================================
+ Hits        30962    30970       +8     
- Misses      12742    12750       +8

Impacted Files	Coverage Δ
src/base/application.h	75% <ø> (ø)	⬆️
src/oneD/Sim1D.cpp	71.72% <0%> (ø)	⬆️
src/transport/MMCollisionInt.cpp	84.44% <0%> (-1.92%)	⬇️
src/oneD/StFlow.cpp	86.94% <0%> (ø)	⬆️
src/thermo/PDSS_HKFT.cpp	67.6% <0%> (ø)	⬆️
src/thermo/MixtureFugacityTP.cpp	28.63% <0%> (-0.14%)	⬇️
src/equil/ChemEquil.cpp	77.95% <0%> (-0.11%)	⬇️
src/thermo/RedlichKwongMFTP.cpp	76.48% <0%> (+0.09%)	⬆️
src/equil/vcs_VolPhase.cpp	74.94% <0%> (-0.16%)	⬇️
src/oneD/Domain1D.cpp	62.3% <0%> (ø)	⬆️
... and 7 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 02c433b...a69151d. Read the comment docs.

[speth]

This looks pretty good to me. I don't think it's a full fix for #683 yet, unless you also want to make the appropriate modifications to the Chemkin conversion scripts.

I would like to avoid adding Boost.Log to our dependencies -- While I generally like Boost, this is unfortunately one of parts that has a compiled library component, and my experience has been that these are very messy dependencies to deal with when you need to support multiple operating systems and a range of different Boost versions. However, a method for raising these warnings through the Python warnings module would be interesting as a future extension.

[speth]

I'm wondering if it would look better to format the warning message without quotes around the method name.

[ischoegl]

No problem. Also suggesting CanteraWarning to make things a little clearer

[speth]

👍 on CanteraWarning

[speth]

It feels a little bit like cheating if you have to modify the output before a warning is issued to get it formatted as desired. One possibility to make multiline warnings more readable would be to automatically add a newline before and after a warning message if it already contains a newline.

[ischoegl]

I added the extra newline here to remain consistent with the existing output, so this is pretty problem specific. I really don't like adding newlines before ... it's up to the user to format things as needed after.

[ischoegl]

@speth ... thanks for the review! Will make the changes which won’t take long.

a method for raising these warnings through the Python warnings module would be interesting as a future extension.

I’ve been wondering about this also, but cannot think of an elegant solution. Definitely another PR

[ischoegl]

One final comment for the record: I noticed that the following warnings depend on a log level. I don't want to change it without discussion as this PR should not change the overall behavior.

cantera/src/transport/MMCollisionInt.cpp

Lines 473 to 493 in 6e60f2f

	if (m_loglevel > 2) {
	writelogf("\nT* fit at delta* = %.6g\n", deltastar);
	writelog("astar = [" + vec2str(vector_fp(a, a+degree+1))+ "]\n");
	if (rmserr > 0.01) {
	warn_user("MMCollisionInt::fit",
	"RMS error = {:12.6g} for A* fit", rmserr);
	}
	writelog("bstar = [" + vec2str(vector_fp(b, b+degree+1))+ "]\n");
	if (rmserr > 0.01) {
	warn_user("MMCollisionInt::fit",
	"RMS error = {:12.6g} for B* fit", rmserr);
	}
	writelog("cstar = [" + vec2str(vector_fp(c, c+degree+1))+ "]\n");
	if (rmserr > 0.01) {
	warn_user("MMCollisionInt::fit",
	"RMS error = {:12.6g} for C* fit", rmserr);
	}
	}

[speth]

With the change to ck2yaml, I agree that this resolves #683.