On Apr 18, 2019, at 12:31 PM, John Rose <john.r.r...@oracle.com> wrote: > > I think it's hard to make these be more convenient (readable, > editable) than Jim and Brian's rules for stripping. They definitely > have an "opt-in" feel to them, because of their extra overheads.
To clarify this paragraph: I mean that a syntax like the __NWC_BLOB__ has an "opt-in" feel. Jim and Brian's rules for stripping do *not* have such string "opt-in" feel, and they have less overhead. With less overhead comes more ambiguity; that's always the trade-off. But the stripping rule which says "strip the most whitespace possible, but equally from all lines" makes for a notation that is very convenient and natural, perhaps enough so to be "opt-out" instead of "opt-in". I suggest that an easy "opt-out" for stripping would be to throw in an extra escape somewhere that disables the stripping machinery, either partially or totally. The *simplest* way to do this is to replace the first payload space on the first line with the escape \040, which denotes an ASCII space, but cannot be confused with stripped indentation. (Or can it? It shouldn't, although I suspect it does in the present implementation. Let's check the spec to make sure. And what about \u0020??? Yikes. That's a real space under a unicode escape, which is processed differently from an octal escape.) Another easy "opt-out" is to transfer the first line (with retained indentation) to the line containing the open-quote. Another not-so-easy "opt-out" would be allowing the sequence BACKSLASH NEWLINE to be elided, as /bin/sh does. This is not in Java. If added, it create (a) a way to break very long lines without resorting to "+" operators, and (b) a way to opt out of stripping, by placing BACKSLASH NEWLINE immediately after the open-quote. And so on… there are lots of ways to skin the cat if we allow ourselves to give him new appendages. (And consider a toroidal cat…) — John
