US_LETTER false negatives

[yakra]

widen the net only enough to catch:

nc.us001:US1AWak_S
nc.us001:US1AWak_N
nc.us001:US1AFra_S
nc.us001:US1AFra_N

[yakra]

if re.fullmatch('\*?US[0-9]+[AB]', w.label) \
or re.fullmatch('\*?US[0-9]+[AB][/_(].*', w.label) \
or re.fullmatch('\*?US[0-9]+[AB][A-Z][a-z]{2}', w.label) \
or re.fullmatch('\*?US[0-9]+[AB][A-Z][a-z]{2}[/_(].*', w.label):
 datacheckerrors.append(DatacheckEntry(r,[w.label],'US_LETTER'))

is nice and clean and readable and we can tell what's going on relatively easily. But it spends a lot of time rechecking long strings at the beginning of the label that we already know are the same.

m = re.search('^\*?US[0-9]+[AB]', w.label)
if m:
 c = m.end(0)
 if c == len(w.label) or w.label[c] in '/_(':
 datacheckerrors.append(DatacheckEntry(r,[w.label],'US_LETTER'))
 else:
 # is it followed by a city abbrev?
 m = re.search('^[A-Z][a-z]{2}', w.label[c:])
 if m:
 c += m.end(0)
 if c == len(w.label) or w.label[c] in '/_(':
 datacheckerrors.append(DatacheckEntry(r,[w.label],'US_LETTER'))

runs over 3 times as fast.

---

On a side note, I really miss the null terminator that C and C++ use; it comes in handy. I'd have preferred to not use the 2nd re.search, but I wanted to avoid a bunch of checking whether array indices overrun the string length. Yes, sometimes I will sacrifice a bit of speed in favor of some brevity of code. :)
Edit: More experimentation... Looks like we can get away with it in this case, because of the checks for c == len(w.label). Invalid indices when slicing will just produce an empty string. Huzzah!

[yakra]

Putting way more time & effort into this than is sane.

box	v.8	v.8A	comments
noreaster	320 334	312 329	Only ran 200 passes. Results multiplied by 10.
lab1	191.3 198.1	197.9 195.5
lab1.5	161.9 158.6	158.0 161.3
lab2	231.5 240.4	226.7 238.5
lab3	341.7 339.9	334.1 336.2	Python 2
BiggaTomato	317.2 307.2	320.9 315.2
epoch	394.4 419.3	423 434.1

I don't need to worry about checking my array indices manually if I enclose the whole thing within try:/except IndexError: Dropping regexes altogether in favor of a more C-like approach that examines the individual characters runs 3-4x as fast depending on what machine.

try:
 c = 3 if w.label[0] == '*' else 2
 if w.label[c-2] == 'U' and w.label[c-1] == 'S' and w.label[c].isdigit():
 while w.label[c].isdigit():
 c += 1
 if w.label[c] in 'AB':
 c += 1
 if c == len(w.label) or w.label[c] in '/_(':
 datacheckerrors.append(DatacheckEntry(r,[w.label],'US_LETTER'))
 # is it followed by a city abbrev?
 elif w.label[c].isupper() and w.label[c+1].islower() and w.label[c+2].islower() \
 and (c+3 == len(w.label) or w.label[c+3] in '/_('):
 datacheckerrors.append(DatacheckEntry(r,[w.label],'US_LETTER'))
except IndexError:
 pass

How much time will this actually save per update? About 0.3s! 🤣