Regex is stripping out word-final and word-initial accented characters as though they are non-word

Bug Reports
1

Bug report here, since there’s no option for a bug report in the topic dropdown.

I have a 3.10 Python going in Anvil, haven’t tried this with not-3.10. The below is a client-side set of code ( (we will also see below where I found out what’s actually going on, but it’s worth showing this.)

As I am a new user, I can only upload one image which is a pain so I’ll type stuff out here.
client-side code in a form:

  print(re.sub(r'^\W+|\W+$', '', "últimas", re.UNICODE))
  print(re.sub(r'^\W+|\W+$', '', "niño", re.UNICODE))
  print(re.sub(r'^\W+|\W+$', '', "vivió", re.UNICODE))

results in

ltimas
niño
vivi

Pretty cut and dried. Here’s how that same code works on my local computer:

últimas
niño
vivió

I typed these characters manually (no copying) with my keyboard, and here’s the unicode inspector results for those characters (proving they’re normal):

image
image2522×1328 213 KB

This is decidedly not how the regex module is supposed to work, and should be investigated!

It DOES work correctly if called from server-side.

Without knowing the specifics, I think [OFF-TOPIC] - Regex - strip leading non-numeric chars was on to something and saw the link to [Done] Re.sub in client side - #2 by stefano.menci ; my ultimate request here is either making it clear that the re(gex) module imported client-side does not work correctly, or forbid its usage, or something.

2

Hi - thanks for reporting.

What you’re seeing here is the difference between javascript and python regex engines, which is frustrating.

For context take a look at the info here which describes how javascript interprets the \W character:

TLDR - \W in javascript is all non alphanumeric characters.

Thanks for the report and we’ll look at fixing that so that \W gets transformed as expected.

The following should work client side

# \u0020 is space character
re.sub(r'$[^\u0000-\u0019\u0021-\uFFFF]$', '', "últimas", re.UNICODE)
1 Like