Incremental multi-level input processing with left-fold enumerator:
	    predictable, high-performance, safe, and elegant
  A practical alternative to Handle and Lazy IO for input processing

	Non-solution: Handle and Lazy IO

GHCBufferIO.hs		Using the conventional Handle IO and Lazy IO
		for part 1 of the running example: reading headers,
		sequences of lines terminated by the empty line

	Non-solution: Manual buffering

BufferNaive.hs		Reading lines from an explicitly allocated and
		filled buffer. Although the code is fairly efficient, 
		it is a burden to write. A sole CR as line terminator
		is not handled. The code is complex already.
		This is the point: a more scalable solution is required.


	Incremental input parsers, processors and transformers

Iteratee.hs		Pure Iteratees

IterateeM.hs		Monadic and General Iteratees
			The main library

IterateeMCPS.hs		The CPS version of the above, which is in some sense
			simpler, making monadic Iteratees look pure


	Sample applications and the comparison with Lazy IO
GetContentsLess.hs	Emulating typical getContents code without 
			any unsafe operations

Wc.hs			Contrasting Lazy and Iteratee IO on the example
			of Unix wc: counting lines and words in a file
			and in a sequence of files


	Random and Binary IO
RandomIO.hs		Random and Binary IO with Iteratee

Tiff.hs			TIFF library and its sample application:
			an extensive example of random and binary IO
			and on-demand incremental processing with iteratees


LowLevelIO.hs		Low-level IO operations 
		These operations are either missing from the GHC run-time 
		library, or implemented suboptimally or heavy-handedly


	Sample input test files

	Test files for part 1 of the running example: reading headers,
        the sequence of lines terminated by the empty line. 
	Each line is terminated by CR, LF, or CRLF.
	We should return the headers in order. In the case of error,
	we should return the headers read so far and the description of 
	the error.
test1.txt		all lines are always terminated by CRLF.
			Set the buffer size is set to 5,
			to ensure case coverage with respect to CRLF
			and the buffer boundary

test2.txt 		lines are variously terminated by CR, LF, or CRLF

test3.txt		The version of test1.txt without the empty line


	Test files for both parts of the running example: reading headers
	and reading lines from the chunk-encoded content
test_full1.txt		Headers plus correctly chunk-encoded content;
			the last line has not terminated
test_full3.txt		Headers plus correctly chunk-encoded content
			the last line is NL-terminated
test_full2.txt		Chunk framing error on the second chunk


	Test files for binary and random IO
test4.txt		Sample binary numbers

gnu-head-sm.tif	        A sample file for TIFF processing
			The image is a GNU logo (from http://www.gnu.org)
			(converted from JPEG to TIFF). Copyleft by GNU.


January 1, 2010.