Add logic to normalize comma-delimited decimals

[ChanceNCounter]

Closes #65

All languages, including English, will now normalize 34,5 to 34.5 before beginning to extract numbers.

Decimal markers can now be specified through extract_number() and extract_numbers() function calls, using a new parameter decimal='.'

Note that these functions will only normalize decimals if they are called as such. Individual parsers, such as extractnumber_es(), have not been modified in any way, but will produce correct output when called via extract_number(lang='es', decimal=',')

For those who don't speak regex, though I encourage you to run the regex through a regex tester, it means this:

\b\d+,{1}\d+\b:
\b a word boundary
\d+ any number of digits, followed by
,{1} exactly one comma, followed by
\d+ any number of digits
\b another word boundary

[ChanceNCounter]

run test

[ChanceNCounter]

A few minutes ago, I realized the way I was trying to distinguish comma-delimited decimals from comma-delimited thousands wasn't gonna work.

So I came here to mark this WIP, and I now see that commit didn't make it to the branch in any case 😳

[ChanceNCounter]

It occurs to me that this argument will almost always be used to parse ',' as a decimal point, and that users who want ',' for decimal points will almost always want '.' for thousands separators.

Should this PR address that?

If so, should it be an additional keyword, thousands=',', or should the decimal keyword be scrapped in favor of actual keywords?

Although the function signatures may become bloated, I'm partial toward two keywords. Standards exist other than full stops or commas for decimal separators. Indeed, it might be worth handling spaces, if specifically indicated in the function call, but that's another layer of complexity, probably involving a while loop.

[JarbasAl]

in a computer a decimal number is always represented with a . regardless of language.

I don't think we will get any stt transcription ever where this isn't the case

It's still good to think about this, chat usage is also important !

[ChanceNCounter]

in a computer a decimal number is always represented with a . regardless of language.

I don't think we will get any stt transcription ever where this isn't the case

It's still good to think about this, chat usage is also important !

Exactly. This PR is so that people can parse data written the opposite way. @TheLastProject asked and shall receive.

[ChanceNCounter]

Alright, here we go, now that I'm free and at a desktop. Say you wanted TTS to read lines from a document, containing yearly something, and you knew the file's formatting: YYYY: #,#

With this PR, you could do:

>>> foo = parse.extract_numbers("1942: 4,5".replace(":",""), decimal=",")
>>> foo
[1942.0, 4.5]
>>> d = format.nice_year(datetime(year=int(foo[0]), month=1, day=1))
>>> d
'nineteen forty two'
>>> bar = d + ", " + format.pronounce_number(foo[1])
>>> bar
'nineteen forty two, four point five'

Pass that along to TTS and you've got some nice behavior.