On Thu, 19 Jan 2012, Hiroki Sato wrote:

The suggestion of 	 is interesting. The problem is that tidy is changing tabs to spaces while still reading the file, when it really should be treating the tab as a special entity while processings tags.

So preprocessing still might be a way to preserve tabs. Just convert them all to 	 before tidy has a chance to change them.

perl -i -pe 's/\t/	/g' book.html

sed works also, but beware of the binary tab necessary for sed's paleolithic regexes:

sed -i -e 's/ /\	/g' book.html

Initial testing shows this does seem to work. The output has genuine tabs which display fine in Firefox but ought to be tested in other browsers. Note that there are some invisible tabs in the HTML output that come from the SGML source. For example, the copyright notice in the Porter's Handbook:

<holder role="mailto:do...@FreeBSD.org">The FreeBSD Documentation Project</holder>

That leading tab on the second line is in the HTML file. Still, replacing those invisible tabs with &#09; instead of spaces should render the same.

Finally, before picking up on the idea of tab-as-an-entity, I worked up a patch to www/tidy-devel which uses the magic value of --tab-size 255 to mean "don't replace tabs". Attached, but I think the &#09; is better.

--- src/streamio.c.orig 2008-03-22 15:00:18.000000000 -0600 +++ src/streamio.c 2012-01-20 12:25:58.000000000 -0700 @@ -351,11 +351,18 @@ added = yes; TY_(AddCharToOriginalText)(in, (tchar)c); #endif - in->tabs = tabsize > 0 ? - tabsize - ((in->curcol - 1) % tabsize) - 1 - : 0; - in->curcol++; - c = ' '; + if (tabsize == 255) { + in->curcol++; + c = '\t'; + } + else + { + in->tabs = tabsize > 0 ? + tabsize - ((in->curcol - 1) % tabsize) - 1 + : 0; + in->curcol++; + c = ' '; + } break; }