cylc reg "[\n]\0 \/\/ exit"

[oliver-sanders]

Currently cylc will let you register suites with unsafe characters. My favourites are:

• Empty string
• Backquotes
• New lines, tab, null characters

[dvalters]

(From earlier office discussion)

One option: Python string methods: isalpha(), isdigit(), isalnum() will test a character is alpha/number/alphanumeric. This also works on unicode strings, and (it would appear) on non-latin alphabets characters.

chinese_string = u'中文。'
for char in chinese_string: print char, char.isalpha()
    
# > 中 True
# > 文 True
# > 。 False

# German
german_string = u'„Hallo! Grüß dich 666!“'
for char in german_string: print char, char.isalnum()

# true for all the letters, including non-English ones, false for all others.
# etc...

Example here:

https://github.com/dvalters/Notebooks/blob/master/is_it_a_letter.ipynb

[matthewrmshin]

In theory, we should be able to get re.compile(ur'\w', re.U) to match Unicode word characters. This didn't work in my environment the last time I looked (when I was working to white list the allowed characters for task names). I have just taken a look in my environment, and it seems to be working now! I'll see if I can repeat this success in a different environment, so we can modify the logic for matching task names as well.

[dvalters]

...I couldn't get that regex to work either

[matthewrmshin]

I think we should revisit this one when we port cylc to Python 3 - #1874, when we can be sure that all strings are automatically Unicode.

[matthewrmshin]

See also #51.

[oliver-sanders]

We may want to roll this one into #2979

[matthewrmshin]

Please close. (The originator of this issue can do the honour.)

[oliver-sanders]

Superseded by the more generic #2979