How do I make the desired Regular Expression?

I am trying to use regexprep to remove certain parts of strings that meet a specific criteria. Here are the criteria in the best English I can think of:

Remove the ## if it occurs at the beginning of the first word even if this word is indented by whitespace.

And here is what I have so far. It works for most of the test cases but fails the last two:

%Test string
str = {'##hello','h##ello','  ##hello','hello','####hello',...
  '  ####hello','h## ello','##','h  ##ello','##hello ##hello'}';

%Match at beginning of word and look behind for whitespace
regs = '(?<!\S)\>(##)';
str2 = regexprep(str,regs,'');    

%What I actually want
str3 = {'hello','h##ello','  hello','hello','##hello',...
   '  ##hello','h## ello','','h  ##ello','hello ##hello'}';

%Pretty visualization
ds = dataset(str,str2,str3,'VarNames',{'Input','Actual_Results','Wanted_Results'})

So it works for all of the cases where there is not a word in front of the ##. However if there is text behind the space, the lookbehind doesn't pick this up. Also I can't use the match beginning of line because I want it to be able to handle n-length white space which beginning of line counts.

[Sean de Wolski]

Bonus points if you can guess the application :)

Answer 1

[Matt Fig]

This seems to work for your test case:

str4 = regexprep(str,'^(\s*)(#{2})','$1');

Or, even simpler:

str4 = regexprep(str,'^(\s*)(##)','$1');

[Sean de Wolski]

Nice, Thanks!!

So just to confirm: we match only the first token '$1' to be '##' and look behind '^' to test if 0 or more '*' whitespaces '\s' are found?

[Matt Fig]

I interpret the regexpstr to say:

First match two tokens at the beginning of the input string. The first token is zero or more white spaces, and the second token is two pound signs. Replace the entire match with only the first token.

But to be honest, REGEXP is nebulous to me, and I am always a little surprised when I get it to work.

[Sean de Wolski]

Yes, you're right, thanks for the explanation. That clears up my confusion about why there was no apparent replacement string - it's the token!