Error messages are not always propagated through primer3_error

[msto]

primer3_error() fails to propagate errors when the logged error includes an equals sign.

When this occurs, the line is stored in primer3_results rather than error_lines.

Could we add stricter validation/parsing of the primer3 output, so underlying error messages are visible?

# contents of `primer3_results`
{'/Users/msto/.miniforge-x86/envs/gia_primer_design/bin/primer3_core:libprimer3.cc:5745, assertion (len >': ' 3) failed'}

# corresponding logged line
/Users/msto/.miniforge-x86/envs/gia_primer_design/bin/primer3_core:libprimer3.cc:5745, assertion (len >= 3) failed

prymer/prymer/primer3/primer3.py

Lines 402 to 412 in 8d332df

	def primer3_error(message: str) -> None:
	"""Formats the Primer3 error and raises a ValueError."""
	error_message = f"{message}: "
	# add in any reported PRIMER_ERROR
	if "PRIMER_ERROR" in primer3_results:
	error_message += primer3_results["PRIMER_ERROR"]
	# add in any error lines
	if len(error_lines) > 0:
	error_message += "\n".join(f"\t\t{e}" for e in error_lines)
	# raise the exception now
	raise ValueError(error_message)

prymer/prymer/primer3/primer3.py

Lines 423 to 435 in 8d332df

	if line == "=": # stop when we find the line just "="
	break
	elif line == "": # ignore empty lines
	continue
	elif "=" not in line: # error lines do not have the equals character in them, usually
	error_lines.append(line)
	else: # parse and store the result
	key, value = line.split("=", maxsplit=1)
	# Because Primer3 will emit both the input given and the output generated, we
	# discard the input that is echo'ed back by looking for tags (keys)
	# that do not match any Primer3InputTag
	if not any(key == item.value for item in Primer3InputTag):
	primer3_results[key] = value

[clintval]

I agree! From the Zen of Python:

Errors should never pass silently.

I would like a way to see all logs from Primer3, perhaps by using the logging module and having the important logs set to INFO/WARNING and all the rest DEBUG.

[clintval]

Sister issue for bwa:

• There should be a way to view the logs that bwa emits #41