We did some specs for new main loop.
I/O is slow
print rarely,
print late;
read early too!
temporarily assume:
prepopulate [q{},q{}]xB
requirement: print '--' unless first
side case: if stdout is autoflush, do so also
invariant: match is set iff there is a match in the buffer someplace
our buffer structure is:
a print buffer and B + A + 1 entries
(which are supposed to be line references to not copy)
and we will also include a prefix string
[[prefix(), \$_,] ... ]
struct of prefix() @TBD -- may include \$Filename as well as $., ':' or '-'.
Defer output formatting to printbuf-appender
if in_matched_context, append shift of line buf onto printbuf
else discard shift of line buf
endif
append current_line to line buf
with prefix() struct or whatnot struct'ed with it
if the current_line matches
we set in_matched_context =true
and count_not_matched=0
else
++count_not_matched
and if count_not_matched == B
start print
else
## if(count_not_matched == (A + B + 1))
if(count_not_matched > (A + B + 1)) ## fix off by one
set_separator()
in_matched_context=0; # clear_matched()
end if
end if
###FLUSH BUFF THING -- no flush logic anymore.
we need the logic for when you shift onto the print buffer:
if match is true, we print only iff we can start print
## DROPPED the print flag (formerly thought separate )
## also DROPPED when a+b+1, stop_print()
concern: if we have match, match, match ...
concern: if we have match, line, match, line, match ...
did we solve that?
if EOF (and in_matched_context) , shift everything to print buffer and print
_______________________________________________
Boston-pm mailing list
[email protected]
http://mail.pm.org/mailman/listinfo/boston-pm
