I. Introduction
	
	TYPO3.MAL is a typo finder/spelling checker that takes about 30% of the
time and slightly less memory than TYPO2C.
	TYPO3 is subject to some limits (explained in more detail below):
.	-Do not use on a document longer than about 1400 words (6000 bytes). If
your document is longer, split it up and check each part separately.
.	-The program requires free memory of between about 4k and 5.5k 
depending on the document's length.
.	-For about 10% of the documents the program will return either a Bad 
Subscript (Error=9), Out of String (Error=14) or Out of Memory (Error=7) during
execution. Why, and what to do, are described below.
	TYPO3 looks for two kinds of typos.
	First, it produces a list of words found once in the document that 
appear nowhere else in memory. The theory is that typos (e.g., transpositional 
errors) will occur only once, but good words are likely to appear more than 
once. Some of the words on the list will be fine, but your job in reviewing the
document will be simplified --- typically to reviewing about 1/3 of the words 
in the document. (The list will have the front and back of any word hyphenated 
at the end of a line).
	Second, the program will display a list of doubled words ("Paris in the
the spring").
	Here's a sense of what it takes to run the program:
	
.		FILE 1	FILE 2	FILE 3
.	
.Bytes long	2330	5009	3885
.Words long	 383	1280	1083
.Distinct words	 194	 285	 238
.	FRE(0) (holding 2500 in TYPO2C and 3000 in TYPO3 for non-string)
.TYPO2C		1247	 497	 442
.TYPO3		1569	 744	 689
.	Time to review document being proofed
.TYPO2C(MIN:SEC) 5:19	20:36	12:39
.TYPO3		 3:00	 7:31	 5:16
.	TOTAL TIME UNTIL DISPLAY OF RESULTS
.TYPO2C		49:27	65:04	53:45
.TYPO3		15:11	18:23	16:14
.	
	Your mileage may vary. Use for comparision only. (Done with little else
in memory and version just this.)

	II. Runtime.
	
	The program displays a copyright/liability notice. Then it displays a 
list of files and asks "Which file to check?" You can answer in upper or lower 
case and with or without the ".DO" extension--the program will convert the 
title to upper, and add ".DO" if it's missing.
	If you enter an illegal file name, the program will crash-- it has no 
error correction.  Because of the way the program allocates memory, if it 
crashes here OR FOR ANY REASON, after fixing the problem do *NOT* simply hit F4
[RUN]. Go back to the menu and reeneter from the menu.
	Assuming you've entered a legal name, the screen will clear and numbers
will flash near the left margin on the second line of the display. These 
numbers represent the number of spaces in the document -- a rough estimate of 
the number of words in the document. The starting time will appear near the 
bottom of the screen.
	The program will create arrays based on a formula using this number to 
hold the different words and count their frequency.  If you get an error 
message for "ERROR= 7" (out of memory) or "ERROR=14" (out of string space) here
you have a couple of choices. If you get an Out of Memory error you can 
increase the "3000" in line 1. If you get an Out of String error either 
increase your free memory by killing a file in memory or decreasing "3000" in 
line 1.  If these don't help, cut the document in two with CUT and PASTE.  
Remember to restart after making any of these changes by running from the main 
menu and to empty the paste buffer before running.
	If you don't get an error, the word "WAIT" will switch back and forth 
between normal and reverse video on the line beneath the count described above.
Beneath "WAIT" a new count (of the number of different words) will appear.
	When the program has finished the file to be checked and is moving on 
to search other files, the name of the file being searched will appear on the 
line below the second counter described above and a new time will be printed to
the right of the time already there.  If instead you get "ERROR=9" (Bad 
Subscript), change the "150" in line 5 to "180" or "200" and rerun from the 
menu. If you are checking a file with a name that is the first part of name of 
another file (i.e., "Memo" and "Memo1"), the second file will not be used to 
look for common words. (But "Memo1" will be checked against "Memo2").
	After the last document file is checked, "SEARCHING ROM" will appear in
place of a file name.
	The screen will clear, then say "STANDBY FOR RESULTS", list the ending 
time, and begin to display "unique" words--all in lower case regardless of how 
they appear in the document.  Every seven lines the screen will say "HIT ANY 
KEY TO CONTINUE" and wait for a key.  When finished the screen may say (if 
applicable) "The following words appear to be doubled:" and print a list of 
doubled words. The screen will then say "*****ALL DONE*****", "Break in 29", 
"Ok".  If you want to see the word list again, type "GOTO 26".
	As the program displays unique words, you should write down any that 
are typos. When the program is finished, enter the document from the menu, use 
the F1 [FIND] function, type in the unique typo, and hit enter.  Then fix the 
typo!	
	
	III. The program.
	
	This section is intended only for people who want to understand the 
approach of the program or fiddle with (and not just use) the program.
	Line 0 prints a copyright and liability notice.
	Line 1 allocates all but 3000 free bytes to string use, defines all 
variables through "T" as integers, the rest as strings. It clears the screen, 
displays FILES, assigns the file to be checked to U and X, U's length to M. If 
not in all caps, the program changes the file name to all caps.
	Line 2 adds "DO" to the file name, if needed. Lines 3 and 4 do the 
space (word) count.
	Line 5 uses the space count to guesstimate the number of different 
words in the file. It then divides this number by 10, and creates two arrays 
with 11 rows (don't forget 0) and a number of columns equal to one-tenth the 
estimated number of words. (One array is for words, the other frequencies.)
	ASIDE: This two-dimension array is the key to TYPO3's speed. I am 
indebted to my colleague, Howard Smith, for the idea of using it. The basic 
notion was this - the first thing one does in a real dictionary is go to the 
section based on the first letter. This looks like a 26 row array. (Sorting the
words within each row would be very time consuming, but picking a row isn't.) 
But the array would need to have as many columns as the most frequent first 
letter -- a gross waste of columns in rows for infrequent first letters. TYPO3 
uses ten rows, each with one or more first letters. Based on frequency each row
should be about the same length. There is a general spill over row.
	Back to the program by lines.
	Line 6 opens the file. Line 7 takes the ASCII value of the next 
character in the document and checks it. If a control code it takes another 
character; if a space or carriage return it declares the end of a word and goes
to 9; if not it makes sure all characters are lower case and then goes to line 
8, which adds the character to the word and goes back to 7. (ASCII values 
instead of letters are used where posible to avoid garbage collection).
	Line 9 strips punctuation from the right of a word, line 10 from the 
left.
	Line 11 checks this word against the last, and if the same adds it to 
the doubled word array, U. 
	Line 12 computes which of the ten main rows of the array contains words
that start with the same letter as the new test word.
	Line 13 checks to see if this word is already in that row of the array.
If so, the frequency is upped; if not and the file being processed (see below) 
is the one to be checked, the word is added to the list and the screen count 
upped by line 18. (If the end of the row is reached, words are added to the 
general overflow row (0) which thereafter is also checked before line 18 is 
reached.) 
	 Line 19 alternates the "WAIT". Line 20 checks to make sure its a file,
not rom, being processed; if so it goes to get the next word.
	Line 21 is reached on an End of File situation. It sets the file flag 
from 0 to 1 on the first pass through; else it returns to the calling 
subroutine.
	Lines 22-23 (borrowed from "FILEN" on this SIG)  find other ".DO" files
to use in cross checking. They also make sure the file to be corrected (X) is 
not rechecked.
	Lines 24-25 search rom words. The other file and ROM lines call the 
main word checking part of the program as a subroutine.
	Lines 26-27 print out the unique words. Lines 28-29 print out the 
doubled words.
	Lines 40-42 are the on error routine. Once a minute they sound 7 beeps 
until interrupted by hitting any key, as well as printing the error number (see
your manual) and line.
	
	Good luck!
		&mike&