⟦6a5ee6210⟧

TextFile

	This directory contains the sources for the spell and typo commands.
This file is intended to aid the dictionary maintainer in tuning it
for his local site.

The Dictionary

	The dictionaries for both typo and spell are stored in the directory
`/usr/dict'.  We will not discuss the typo dictionary here as it is not
intended to be extended, rather being provided as a screen of the most common
words.  Remaining words are ordered statistically, rather than by dictionary
lookup.  Conversely, spell uses an exact membership tester rather than one of
the several approximate membership tester techniques found in the literature.
The resultant small speed penalty is more than offset by the ability to have an
almost perfect lookup record with essentially no misses.  The only misses
will be such misspelling as the wrong homonym (`their' for `there') or
the misuse of a proper name (e.g. `manuel' for `manual').

	The dictionary is maintained as a set of word lists.  As there
are really two dictionaries--a British one and an American one--a set
of common words (`common') is maintained as well as files with each of
American (`american') and British (`british') spellings.  More will be
said later about what words should go into each file.  In addition, there
is a file reserved for the local dictionary maintainer, called `local',
which allows him to add words without changing the distributed lists.
These files can be merged (`sort -mu common british local' or
`sort -mu common american local') to produce the two master lists.
These lists are never normally stored in this form, rather the compressed
lists are stored.  This compression reduces the amount of I/O that must
occur for the spell command to read in the dictionary.  These compressed
files are named `clista' and `clistb', for American and British, respectively.
The shell file `build' reconstructs the compressed lists from these components
by using the above `sort' commands and the internal comman
`/usr/src/cmd/spell/spellin' which takes a list of words on its
standard input and produces a compressed list on its standard output.

	In these words lists, a trailing asterisk (`#') on a word implies that
there is an optional `s'.  This may signify a plural or a different verb form.
This shorthand is included so that such plurals can be cheaply maintained
without the sacrifice in accuracy that would occur with arbitrary suffix
and prefix stripping code.

British Versus American

	Since there is some dissension in British spelling regarding words
that end in `ize' or `ise', the `ize' spelling is included in the common
file and a corresponding entry with `ise' is put into the British file
(e.g. optimize vs. optimise).  Such words as `honor' vs. `honour' are
put into the American and British files, respectively.  However,
in the US there are words which are spelled two ways (e.g. `glamour' vs.
`glamor') so that the former is put into the common file and the
latter is put into the American file.  There are no hard and fast rules
in any case, and thus there are bound to be disagreements no matter what
is done here.

Updating Word Lists

	Spell maintains a history file, `spellhist', which contains all
misspelled words that are found.  This will contain a large number of
words that should not go into the dictionary (e.g. opcodes, variable names,
very obscure technical terms, etc.) as well as words that should be added.
This list should be screened periodically and truncated.

	Some of the files, especially `common', are too large to edit with
`ed' and thus the stream editor (`sed') can be used to manipulate it.
Also, if it is desired to change `common' rather than `local', new words
can be put in by creating a new words file, say `new', and merging it
with sort into a new `common' file (e.g. `sort -mu -o common common new').

	As well, it is always a good idea to check when adding new words
that the singular is not already there to save adding a plural.
For example, if the word `macro' was already in the dictionary and
`macros' was to be added, one could issue the following `sed' command
to change `macro' to `macro#' (singular or plural):
	sed 's/^macro$/macro#/' common >new; mv new common

	One last word of caution is in order.  Word lists that are updated
should be scrupulously checked for errors with a dictionary.  Nothing is
worse than a spelling dictionary with misspellings in it.  This can happen
quite easily if a common misspelling is believed from the history file.
DataMuseum.dk

Commodore CBM-900

⟦6a5ee6210⟧ TextFile

Derivation

TextFile
