legacy f77 code is not detected with spaces

[kostyfisik]

This code compiles and works well, however, T H E N end E L S E statements are not detected\colored\folded.

      ALPHA=SQRT(DFLOAT((L-M)*(L+M+1)))/2.D0  
      BETA =SQRT(DFLOAT((L+M)*(L-M+1)))/2.D0  
      IF(ABS(M+1).LE.L)  T H E N  
      I=L*L+L+M+2  
      Z1=YLM(I)  
                         E L S E  
      Z1=CZERO  
                         E N D    I F  
      IF(ABS(M-1).LE.L)  T H E N  
      I=L*L+L+M  
      Z2=YLM(I)  
                         E L S E  
      Z2=CZERO  
                         E N D    I F  
      I=L*L+L+M+1  
      Z3=YLM(I)  
      DLMH(1,II)=COEF*(BETA*CT*CF*Z2-DFLOAT(M)*ST*Z3 

[satamas]

That's won't fix kind of issue. I do know that we don't folow fortran spec in part of spaces, but we've did it intentionally. This language feature is one of those features that's extreamly complex to support by modern lexer/parser and at the same time it's usage should be eliminated from supported code (from my point of view), so marking this as an error is rather reasonable.

[kostyfisik]

Ok. I just though that it can be some kind of pre-parser module, which search for space separated keywords and pass the normalized code to the lexer/parser. This special spaces-handling should probably be presented in top-level readme.

Does it mean that converting this type of code to free form by the plugin may generate a wrong working programm?

[satamas]

I just though that it can be some kind of pre-parser module, which search for space separated keywords and pass the normalized code to the lexer/parser

I think that this example shows that there is no good way of parsing fortran code when spaces are ignored. Also you should take into account that one can use keyword as variable name or split keyword token between lines.

This special spaces-handling should probably be presented in top-level readme.

Yes, sounds reasonable, I'll update readme

Does it mean that converting this type of code to free form by the plugin may generate a wrong working programm?

As far as I see it shouldn't. Conversion depends only on PsiComment elements and it's rather hard to break comment parsing.

[kostyfisik]

See #42 for conversion problem.