The offsets of added parameters are wrong in some cases (not always)

[tsantalis]

In commit tsantalis/RefactoringMiner@5e79668
there are 3 added parameters,
but the offsets are shifted below the actual location of the added parameters

[onewhl]

@tsantalis thank you! @anchouls could you take a look at it, please?

[anchouls]

@tsantalis @onewhl UMLParameter parameter in AddParameterRefactoring class (from RefactoringMiner tool) returns an incorrect LocationInfo, specifically an incorrect offset


Correct startOffset = 16267

The lines and columns correctly

[tsantalis]

@anchouls @onewhl
We get the offset directly from the ASTNode

ASTNode node;
this.startOffset = node.getStartPosition();
this.length = node.getLength();
this.endOffset = startOffset + length;

How is it possible that the line number and column number are correct and the offset is not?
The line number and column are computed based on the offset.

[tsantalis]

Notepad++ gives a

start offset 16614, line 347, column 13
end offset 16643, line 347, column 42

The offsets are closer to what the ASTNode is giving

[anchouls]

@tsantalis @onewhl

The IDEA doesn't count line breaks when counting characters (unlike Notepad++, VS Code, etc.). Usually, the startOffset also does not include line breaks, and highlighting works well. Why line breaks were counted now is currently an open question.

16613 - 16267 = 346 - number of line breaks

[tsantalis]

@anchouls @onewhl
Is this related to the DOS vs. Unix line endings?

DOS uses carriage return and line feed ("\r\n") as a line ending, while Unix uses just line feed ("\n")

This file is using CRLF ("\r\n") endings (2 characters).

[tsantalis]

@anchouls @onewhl

My students told me that IntelliJ ignores the actual line endings of the file, and commits code using the Unix LF ("\n").

There was a file in our repository with CRLF ("\r\n") line endings, and IntelliJ changed everything to LF ("\n") after the commit.

It seems IntelliJ has some preference or configuration for LF ("\n") by default.