Understanding TRegExpr

[pmesquita]

Guys,.
Anyone know TRegex (https://github.com/andgineer/TRegExpr)?

I'm playing with TRegExpr and according to the website: https://regex101.com/ if I use the expression: ^[a-z]([\\.](?![\\.])|[a-z]){5,50}[a-z]$ to validate the text: foo.bar.

The site is valid, but if you throw this same expression to the FPC it "screams" that the position is unknown...

Note: the two backslashes are because the "schema" field is made a "GetJSON" in the content..

Validation code:

Function CheckPattern(AValue, APattern: String): Boolean;
var
   LRegex: TRegExpr;
  Begin
   LRegex:= TRegExpr.Create;
   try
    LRegex.Expression:= APattern;
    LRegex.ModifierI:= False;
    LRegex.ModifierS:= True;
    LRegex.ModifierX:= True;
    Result:= LRegex.Exec(AValue);
   finally
    FreeAndNil(LRegex);
   End;
  End;

[AlexTP]

I am the maintainer.
FPC cannot find this part: (?![\\.])
It is called 'assertion' and it was added to FPC 'trunk' but was not added YET to the last FPC release.

And, 'assertion' must be at the very end of expression.

[pmesquita]

I'm talking to the right person..

understood, but I'm using the master version in Git.
So the expression should look like this: ^[a-z][\\.][a-z]{5.50}(?![\\.])$ ?

[AlexTP]

That is also 'not at the very end', notice the $ at the very end. BTW, you can test this in CudaText (with latest TRegExpr). Editor shows detailed regex errors.

[pmesquita]

what's up..

I think I understand, so the 'assertions' should always be declared or at the beginning or end of the expression would it be a current limitation or implementation?

Because in my understanding and using CudaText I arrived at the expression: ^[a-z]+[\.][a-z]{1,50}$(?![\.]) where it works for what I need, however I a question arose:

Would the assertion (?![\.]) at the end of the text (after the $) be a regular expression for the "[\.]" declared earlier in order to avoid repetition of the dot (.) ?

[AlexTP]

Yes, it is the limitation, https://regex.sorokin.engineer/en/latest/regular_expressions.html#assertions - docs tell it.

Would the assertion (?![\.]) at the end of the text (after the $) be a regular expression for the "[\.]" declared earlier in order to avoid repetition of the dot (.) ?

I did not understand. En is not my native lang.

[Martin_fr]

In this case the regex may be "fixable"

The orig regex matches any text that
- starts and ends in a letter
- is between 7 and 52 chars long
- can have any amount of single dots between any 2 letters.


Since the text has only dots and letters, the lookahead (?!\\.)  means a letter must follow => and hence this is a word boundary.

I did not test, but I would expect that this should work:

Code: Text  [Select][+][-]

1. ^[a-z](\\.\\b|[a-z]){5,50}[a-z]$

Depending on what "only at the end of an expression" means exactly... (I.e. only at the very end of the entire regex, or any sub-expression?), there may be another options.
The condition is that the overall length should be between 7 and 52 chars....
If we temporary ignore the length, the regex becomes real simple: 

Code: Text  [Select][+][-]

1. ^[a-z]+(\\.[a-z]+)*$

- starts with one or more letters
- followed by zero, one or more: dot with at least one letter after it
  (that also ensures the last char is a letter)

If we can place the lookahead at the very start, we can check the length first

Code: Text  [Select][+][-]

1. ^(?=.{7,52}$)[a-z]+(\\.[a-z]+)+$

That does 2 matches, both from the start, and both required to succeed.
- check the length
- check the content as explained