home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Gigarom 1
/
Gigarom Macintosh Archives (Quantum Leap)(CDRM1080320)(1993).iso
/
FILES
/
DTP
/
P-Z
/
Pairing Knife.cpt
/
Pairing Knife д
/
Pairing Knife - Read Me
next >
Wrap
Text File
|
1992-06-11
|
5KB
|
134 lines
RELEASE NOTES FOR PAIRING KNIFE
This is a utility of the 'one dumb job' variety. It does one job, and
does it fast and with minimal hoorah.
Pairing Knife is SYSTEM 7 ONLY. Its only interface is Drag and Drop,
so you cannot use it with earlier Systems. If you double-click on it
from a System 6 machine, you will get an error message and the
software will quit gracefully.
To use Pairing Knife _with_ System 7, simply select the files you
want to analyze and drag them on the program's icon or an alias of
it. A new file called "Pairing Knife.TXT" will be created, and your
original source files will remain unaltered.
What does it do?
Pairing Knife does an unbiased kerning pair frequency analysis of
files of type TEXT. Unbiased means that all printing characters (from
ASCII 32 to 255) are considered in the analysis. Kerning Pairs, of
course, are pairs of characters that can appear in text and that can
have kerning applied between them. Frequency Analysis means that
every pair combination is counted each time it appears.
Pairing Knife writes a report of the relative frequency of kerning
pair combinations to a TEXT file called "Pairing Knife.TXT". You can
use this file to help establish a kerning strategy.
Pairing Knife.TXT issues the report with the pairs sorted in strict
ASCII order. You can use Word or sorting software to sort to another
order. Word will show the list in ascending order of relative
frequency.
Why does frequency analysis help?
As you'll see if you run the software, the slope from least frequent
to most frequent pairs is extreme. Editing the 500 most common pairs
will have a greater effect on the overall appearance of your text than
editing the 500 ugliest pairs - which may not be all that common in
real text. Surely you should go after both, to whatever extent you
can stand, but a knowledge of the relative frequency of a particular
combination can help you make the best use of your time.
Take note of these provisos, however:
A frequency analysis is only as good as the sample text. If you feed
Pairing Knife poetry when your day-to-day work concerns electrical
supplies, you will learn little of real value. Likewise, if your
sample text is filled with garbage characters, the frequency reports
on those pairs will be skewed.
Pairing Knife is treating all instances of tab, return and linefeed
as spaceband characters; all other control characters are ignored.
The first character in a file is presumed to be preceded by a space,
and the last character is presumed to be followed by a space. In no
case, however, will multiple spaces be counted. Thus if you had a
file like this:
a
b
c
å
∫
ç
Pairing Knife would issue this report:
1 a
1 b
1 c
1 å
1 ç
1 ∫
1 a
1 b
1 c
1 å
1 ç
1 ∫
Total Pairs: 12
Toward a finer balance...
I made Pairing Knife unbiased so no one would have to struggle to put
the toothpaste back in the tube. However, a 'biased' analysis is
probably more useful in real life. For that reason, a Torquemada set
called "'Bias' Pairing Knife Sample" is enclosed with this archive.
The set is commented, and you should feel free to adapt it to your
particular circumstances. Only one change bears close scrutiny: all
capitals in the sample are being turned into lower case letters. The
reason for this is that the same one (virtual) kerning pair can turn
up in three permutaions (for example, 'aw' can show up as 'AW', 'Aw'
or 'aw'). In reality, if _any_ of those turn up a lot in the sample,
then _all_ of them will turn up a lot in real text. In consequence,
you would want to kern all three, even if 'AW' is relatively rare in
the report. But an unbiased report might fail to make manifest the
true frequency of the (virtual) pair (600 is more noteworthy than 150
+ 200 + 250). So converting everything to lowercase reveals
information that might otherwise be obscured.
Ideally, a sample should be _enormous_. It should include all the
text you can lay hands on, with the stipulation that that text
actually reflects your real, day-to-day needs. It should be biased
to reflect your actual working conditions (e.g., omit numbers if you
have to have tabular alignment). This is a lot of work...
Here's a way of doing it fast:
* Accumulate _copies_ of your text files, all in one folder
* Drag & Drop them onto Cat o' Seven Tails
* Drag & Drop "Concatenated files…" plus your 'biasing' Torquemada
set(s) onto Torquemada
* Drag & Drop "Concatenated files….TQM" onto Pairing Knife
* Drag & Drop "Pairing Knife.TXT" onto Word and select Utilites Sort
* Pass the list off on the designated KerningDrone and go home
early (grin)
That's it... Not the coolest thing I've ever written, but it's a fun
toy with a nice, if short-lived, benefit.
Very Best,
Greg Swann
6/10/92