Semantic Equivalence (removing try/excepts from Measurement)

[samiebee43]

Critical sections of code tend to become the hardest to maintain. The importance of the area makes it hard to justify refactoring that might introduce bugs. The centrality of these sections tend to make them frequent targets of added complexity.

Here I will accumulate justifications and proofs related to code simplifications.

#199 is a clear example of this and the starting point for this discussion post.

[samiebee43]

I would like to remove the try/except pairs on 860 and 871.

ENLIGHTEN/enlighten/measurement/Measurement.py

Lines 859 to 874 in fd5f071

	for field in self.get_extra_header_fields():
	try:
	md[field] = self.get_metadata(field)
	except:
	# get_all_metadata is used to generate a tooltip and field "ccd c0" may yield an index error
	md[field] = "<missing>"
	for field in Measurement.CSV_HEADER_FIELDS:
	if field not in Measurement.ROW_ONLY_FIELDS:
	if field == "Timestamp":
	md["Timestamp"] = self.timestamp.strftime("%Y-%m-%d %H:%M:%S.%f")
	else:
	try:
	md[field] = self.get_metadata(field)
	except:
	md[field] = "<missing>"

Starting with 860.

We have assignment to md[field] in either case, so semantics associated with dictionary assignment (parameter shadowing for instance) do not need to be considered.

The try/except can be removed iff self.get_metadata does not raise an exception for all fields in the domain of self.get_extra_header_fields(). The domain is easily determined since that function resolves to a shallow copy of a constant. Defined on line 210.

I checked manually that all members of that domain have an associated if/return, and so would not return None (potentially breaking some code downstream of get_metadata).

This section could raise an exception if process_reading or plugin_metadata is undefined--or if pm is not a dictionary-like. Ignoring this for now.

ENLIGHTEN/enlighten/measurement/Measurement.py

Lines 774 to 782 in fd5f071

	def get_metadata(self, field):
	field = field.lower()
	# allow plugins to stomp metadata
	if self.processed_reading.plugin_metadata is not None:
	pm = self.processed_reading.plugin_metadata
	for k, v in pm.items():
	if field == k.lower():
	return v

The following call will not exception, but it may return None.

ENLIGHTEN/enlighten/measurement/Measurement.py

Line 784 in fd5f071

wavecal = self.settings.get_wavecal_coeffs()

If it did return None, the following sections would exception 'Nonetype has no len'. The method can be easily adjusted to handle all branches to avoid this.

ENLIGHTEN/enlighten/measurement/Measurement.py

Lines 801 to 805 in fd5f071

	if field == "ccd c0": return 0 if len(wavecal) < 1 else wavecal[0]
	if field == "ccd c1": return 0 if len(wavecal) < 2 else wavecal[1]
	if field == "ccd c2": return 0 if len(wavecal) < 3 else wavecal[2]
	if field == "ccd c3": return 0 if len(wavecal) < 4 else wavecal[3]
	if field == "ccd c4": return 0 if len(wavecal) < 5 else wavecal[4]

Next areas of interest will be function calls generated by get_metadata.

• self.settings.full_model()
• self.settings.is_sig()
• self.settings.excitation()
• self.settings.pixels()
• self.processed_reading.is_cropped()
  (leaving out call to str, which rarely throws exceptions)

self.settings is a SpectrometerSettings instance from Wasatch.PY. These accessors are copies of data acquired though device communication. Themselves they are not not prone to pyusb related device exceptions.

In the interest of scope, I'm leaving out the possibilities of certain intermediate expressions being undefined, such as self.save_options.

The one on 871, overlaps, with just a different subset of the domain.

Overall, it's unlikely either of these sections will incur unhandled exceptions due to the removal of these try/excepts.

[samiebee43]

Turns out this was only used the one time