Re: char_width('\t'), no, the question is not about offsets

[David T. <dv...@da...>]

>> is it your reading of the specification that writing the #\Tab
>> character to the output stream determinately keeps the output-stream 
>> at
>> the start of a line?
>
> No. But any other value would be just as wrong. 8 would be wrong on 
> Windows,
> 4 or 2 would be wrong on Unix. And it depends on the settings of the 
> terminal
> (xterm or so).
>
> FRESH-LINE just tests the whether the value is 0. But the 
> pretty-printer
> is highly interested in the magnitude, and since you cannot get this 
> right,
> it doesn't really matter what FRESH-LINE does in this context.

I do not advocate computing the offset when the tabulation character is 
written to the output. But fresh-line should NOT look at line-column at 
all; instead, it should look at a bit that indicates that the line is 
not at its start, and #\Tab should be among characters that set the 
bit.

> I'd recommend to just avoid the tabs. FORMAT ~nT outputs enough spaces 
> to
> get to column n; therefore you don't need tabs.

If CLISP supports recognizes #\Tab, it should handle it in accordance 
with specification. The question is not whether something is good or 
bad; but whether FRESH-LINE is conforming or not. I believe that 
CLISP's fresh-line implementation is not conforming.

Imagine, I deal with a text format that uses #\Tab to denote 
indentation levels; spaces are a part of content and do not count. 
Therefore, I do need tabulation characters.

But this is not important, what I am trying to say is that FRESH-LINE 
in CLISP does not conform to the Common Lisp specification, because it 
does not output #\Newline on fresh-line when it cannot determine 
whether the stream is at the start position of a line.

A useful approach would be to set it char_width to -1, and to not 
increment it when it is -1, meaning that the position cannot be 
determined. Another way is to allocate a separate bit field just to 
mark that a line is 'dirty' (and I am not sure that the latter is 
cleaner than the former).

David Tolpin