Adding an advanced data structure for Title/Word in the CaseChangers …

[simonharrer]

…implementation.

@koppor this is what I ment. What do you think?

[koppor]

While trying to make test cases, I discovered net.sf.jabref.bst.BibtexCaseChangersTest.testChangeCase(), which seems to cover most of our wishes and even can handle intra word casings correctly: Vall{\\'e}e -> VALL{\\'E}E. HAllo {worLD}. HOW -> hallo {worLD}. how.

I would suggest to remove CaseChangers completely and to use net.sf.jabref.bst.BibtexCaseChanger only to avoid code redundancy. I am aware that the code of BibtexCaseChanger is more ugly than the current code of CaseChangers, but it covers the strange cases (e.g., case also LaTeX characters).

[simonharrer]

I will have a look. But the bst package is a candidate to be thrown away as
it is unmaintainable.

[simonharrer]

I had a look: The CaseChangers have Title, Upper, Lower, UpperEachFirst and UpperEach whereas the BibtexCaseChangers only have Title, Upper and Lower. Small words are not detected in the BibtexCaseChangers algorithm. The only thing that the BibtexCaseChangers are better in is the detection of special latex characters like umlauts.

From my perspective, the BibtexCaseChangers class is hard to maintain. And it is used heavily in the bst package, making it very hard to improve the class as its interface has to stay the same. I think the current solution is good enough for a lot of use cases already. We would loose a lot and gain too little if we would replace the CaseChangers by the BibtexCaseChangers.

I would suggest to leave the BibtexCaseChangers class alone, and in the future remove the bst package and replace it with something else. Or just delete it. ;)

If someone has really a lot of time, they can use the test cases to drive an improved CaseChangers version. One simply has to determine a different boolean[] protectedChars array based on the title, namely, to not make every char within curly braces protected.

[koppor]

I made the interface a bit more modern (using enum instead of char as parameter). Still procedural style than object oriented, but for that use case, I think procedural style is more fitting. 😇

I made test cases based on #176 (comment) at

jabref/src/test/java/net/sf/jabref/bst/BibtexCaseChangersTest.java

Line 107 in 03703a5

public void testTitleCase() {

. Think, it's time for a call.