How to include a literal '#' in a Tatsu grammar?

I can't get Tatsu to parse a grammar that includes a literal '#'.

Here is a minimal example:

G = r'''
atom = /[0-9]+/
     | '#' atom
     ;
'''

p = tatsu.compile(G)
p.parse('#345', trace=True)

The parse throws a FailedParse exception. The trace seems to show that the parser is not matching the '#' literal:

<atom ~1:1
#345
!'' /[0-9]+/
!'#' 
!atom ~1:1
#345

If I change the grammar to use a symbol other than '#', it works fine. For example this works:

G1 = r'''
atom = /[0-9]+/
     | '@' atom
     ;
'''

tatsu.parse(G1, '@345')     --> ['@', '345']

Unfortunately, I can't change the format of the input data.

This is likely a bug in the version of TatSu you are using.

If you need to stick to that version, please try including @@eol_comments :: // or a similar pattern in the grammar.

This works for me:

[ins] In [1]: import tatsu                                                                                      

[ins] In [2]: G = r''' 
         ...: atom = /[0-9]+/ 
         ...:      | '#' atom 
         ...:      ; 
         ...: ''' 
         ...:  
         ...: p = tatsu.compile(G) 
         ...: p.parse('#345', trace=True)                                                                       
↙atom ~1:1
#345
≢'' /[0-9]+/
#345
≡'#' 
345
↙atom↙atom ~1:2
345
≡'345' /[0-9]+/
≡atom↙atom 
≡atom 
Out[2]: ('#', '345')

AFTERNOTE: Yes, the above output is from the master version of TatSu (sequences return tuple), but I just checked against v4.4.0, and it's equivalent.