infotype

#!/usr/bin/awk -f

# Analyze a Markdown (MDITA) file
# and tag it with a topic type
# based on its content. Patterns
# and weights can be adjusted to suit.
#
# Usage: awk -f [me] [-v var=value] file.md
#	var:
#		DEBUG - spit out debug msgs if true (0)
#		generic - default topic type (topic)

BEGIN {
	# set up the weight table (see end of script)
	numtests = weight_setup();

	if( !generic ) generic = ""; # default type if no clear winner (blank means type "topic")
	if( !DEBUG ) DEBUG = 0;
}

FILENAME != oldfile { # new file incoming
	type = get_weight();
	if( !type ) type = generic;
	if( type && oldfile != "" ) {
		sysstr = parse_hdg( hdg1, oldfile, type );
		if( DEBUG ) print "###parse_hdg returns:", sysstr;
		if( sysstr ) system( sysstr );
	}
	hdg1 = "";
	delete results;
	delete tests;
	oldfile = FILENAME;
}

/^# / {
	# save the heading (but let the pipeline test it)
	hdg1 = $0;
}

{
	# run tests on each line
	# skip tests that have already passed
	for( i=1; i<=numtests; ++i ) {
		if( !tests[i] && $0 ~ weight[i,"pattern"] ) {
			results[weight[i,"topictype"]] += weight[i,"weight"];
			tests[i] = 1; # mark test as passed
		}
	}
	next;
}

END {
	type = get_weight();
	if( !type ) type = generic;
	if( type && oldfile ) {
		sysstr = parse_hdg( hdg1, oldfile, type );
		if( DEBUG ) print "###parse_hdg returns:", sysstr;
		if( sysstr ) system( sysstr );
	}
}

# # # # functions follow # # # #

# returns the topic type with the highest score
# return can be blank, in which case use the generic global
# ties go to the first type to be checked
function get_weight(   win, hiscore, t) {
	# uses global results array
	win = "";
	hiscore = 0;
	for( t in results ) {
		if( results[t] > hiscore ) {
			win = t;
			hiscore = results[t];
		}
	}
	if( DEBUG ) print "###get_weight: hiscore=" hiscore "; type=" win;
	return win;
}


# returns a command line to add the proper topic type to the heading
# return can be blank, in which case don't do anything
function parse_hdg( hd, fn, newtype,   str ) {
	# skip if the heading already has a topic type or outputclass
	if( hd ~ /{.*\.[a-z]/ )
		return "";

	# check for ID construct: {#idname}
	if( hd ~ /{.*#[a-zA-Z]/ ) {
		str = "sed -i -e '/^# /s/{/{." newtype " /' " fn;
		return str;
	}

	# default, no ID/outputclass
	str = "sed -i -e '/^# /s/$/ {." newtype "}/' " fn;
	return str;
}
			
# You can add your own rules here
function weight_setup(   w ) {
	# format: pattern, topic type, weight
	# returns: number of entries

	w = 1;

	# bullet list = concept (maybe)
	weight[w,"pattern"] = "^[ \t]*\\* +";
	weight[w,"topictype"] = "concept";
	weight[w,"weight"] = 0.5;
	++w;

	# numlist = task
	weight[w,"pattern"] = "^[ \t]*[0-9]+\\. +";
	weight[w,"topictype"] = "task";
	weight[w,"weight"] = 1.5;
	++w;
	
	# first word of title ending in "ing" = task
	# doesn't always seem to help
	# weight[w,"pattern"] = "^# +[^ ]*ing ";
	# weight[w,"topictype"] = "task";
	# weight[w,"weight"] = 0.5;
	# ++w;
	
	# deflist = reference
	weight[w,"pattern"] = "^ *:  *";
	weight[w,"topictype"] = "reference";
	weight[w,"weight"] = 1.0;
	++w;
	
	# table = lean toward reference
	weight[w,"pattern"] = "^\\| ";
	weight[w,"topictype"] = "reference";
	weight[w,"weight"] = 0.5;
	++w;
	
	# tables might be HTML
	weight[w,"pattern"] = "<table";
	weight[w,"topictype"] = "reference";
	weight[w,"weight"] = 0.5;
	++w;
	
	# "About" in the heading is usually a concept
	weight[w,"pattern"] = "^# .*[Aa]bout";
	weight[w,"topictype"] = "concept";
	weight[w,"weight"] = 1.5;
	++w;
	
	# "Commands" in the heading is usually a reference
	weight[w,"pattern"] = "^# .*[Cc]ommands";
	weight[w,"topictype"] = "reference";
	weight[w,"weight"] = 1.5;
	++w;

	# "Syntax" at the beginning of a paragraph is prob. a reference
	weight[w,"pattern"] = "^Syntax:";
	weight[w,"topictype"] = "reference";
	weight[w,"weight"] = 1.0;
	++w;

	if( DEBUG ) print "###weight_setup:", w, "tests read.";
	return w;
}