ENB: An alternative to g.splitLines

[Edward K. Ream]

The fix to a bug in Leo's Rust importer requires an alternative to g.splitLines:

def splitLines(s: str) -> list[str]:

    return s.splitlines(True) if s else []

The problem arises from str.splitlines. Importers should split lines only newlines. A crucial invariant fails otherwise.

A new function is required: g.splitLinesAtNewline. This function splits lines only at newlines. It does not split lines at form-feeds and other unusual line-ending characters. 

This function is surprisingly complicated. I had to develop it using a new unit test.

Summary

All of Leo's importers use g.splitLinesAtNewline.

g.splitLines will remain as it is: any change would be a breaking change to Leo's API.

Edward

P.S. Here is g.splitLinesAtNewline:

def splitLinesAtNewline(s: str) -> list[str]:
    """
    Split lines *only* at '\n', preserving form-feeds and other unusual line-ending characters.
    """
    if not s:
        return []
    lines = s.split(sep='\n')
    if lines[-1] == '':
        lines.pop()
    lines = [f"{z}\n" for z in lines]
    if not s.endswith('\n'):
        lines[-1] = lines[-1][:-1]
    return lines

Notes

The "s" arg must be a unicode string. byte args are not allowed, but all tests pass with g.splitLines defined this way:

def splitLines(s: str) -> list[str]:
    return splitLinesAtNewline(g.toUnicode(s))

To repeat, I am not going to change g.splitLines!

EKR