[flexichain-devel] A couple of flexichain questions

[Robert Strandh]

Brad Beveridge writes:
 > I've thought about the two methods (gap buffer vs linked list of
 > lines), and basically the performance for either method is probably
 > about the same for most operations.  

Probably.  However, a list of strings will have very bad performance
if you want to edit code with very long lines.  

My intention for Climacs was to eventually use a tree of flexichains.

 > The one big place that a
 > gapbuffer wins is when you want to treat a large section of the buffer
 > as a read-only string.  For a list of lines you need to cons up a new
 > string, for a gap buffer (assuming it is locked), you simply need to
 > move the gap and take a pointer into the gap buffer.

Sure. 

 > I envision using this "large string" mode of the gap buffer for:
 >   - searching one huge string with CL-PPCRE
 >  - saving the buffer in one call
 > 
 > So after all my rambling I have two questions:
 > 1) Is it appropriate to use Flexichain in this "one large string"
 > manner?  Obviously you don't want to always convert to this form, but
 > it is handy for running a search on the buffer and some other uses.

I think it would work pretty well. 

 > 2) What is involved in creating a function that consolidates the data
 > into a contiguous chunk that is suitable for creating a displaced
 > array into?

Well, there is an additional complication.  A Flexichain is circular,
so its logical beginning could be in the middle of the physical
array.  

Aside from that, you just have to move the gap to the right place, I
would think. 

-- 
Robert Strandh

---------------------------------------------------------------------
Greenspun's Tenth Rule of Programming: any sufficiently complicated C
or Fortran program contains an ad hoc informally-specified bug-ridden
slow implementation of half of Common Lisp.
---------------------------------------------------------------------