feature: string pattern anonymizer, build complex strings by fetching values from other anonymizers

[pounard]

See documentation for more information.

TL;DR:

anonymization:
    default:
        customer:
            biography:
                anonymizer: pattern
                options: {value: '{firstname} {lastname} lives in {address:locality} in {address:country}, he has [2,7] cats.'}

String range

By adding [MIN,MAX] anywhere in your text, it will be replaced by a random integer
in the given range. For example "I want [7-111] apples".

You can use negative integers such as [-10,10] or even a fully negative integer
range [-127,-34].

Single value from another anonymizer

By adding {ANONYMIZER} anywhere in your text, it will be replaced by a random value
given by the anonymizer. For example "My email address is {email}, what's your ?".

Column value from another anonymizer

By adding {ANONYMIZER:COLUMN} anywhere in your text, it will be replaced by a random
value for the named column given by the anonymizer. For example, "The country I live into is {address:country}.".

If you use more than one column of the same anonymizer in the same text value, the same
generated row from the target anonymizer will be used. For example, this text will give
a consistent result: "The country I live in {address:locality} in {address:country}.".

[SimonMellerin]

Seems really good !

Maybe an alternative could be to accept this kind of conf

anonymization:
    default:
        customer:
            biography:
                anonymizer: pattern
                options: 
                     value: '{1} {2} has email address {3}.'
                     1: firstname
                     2: lastname
                     3: 
                         anonymizer: email
                         options: {domain: 'custom-domain.com'}

This can be an addition for advanced cases. what do you think ?

But this is may be far too complex for a first version.

[pounard]

Maybe an alternative could be to accept this kind of conf

It seems vastly different from what I have done, but it's definitely an interesting idea. How about opening an issue to add this dialect variant as well ?