symvars

[Ken Irving]

I've been using non-resolving symbolic links to store string values,
and have found them to be quite useful. One use is as named variables
on the filesystem, i.e. symbolic link variables or symvars, also with
the convention of having the prefix '=' on the value.

Here's an example listing of some symvars describing aspects of an
application object in a directory:

$ ls -og
...
lrwxrwxrwx 1 9 Jan 24 10:41 dcpid -> =326AF6FE
lrwxrwxrwx 1 11 Jan 24 10:42 lat -> =70d 16.23m
lrwxrwxrwx 1 13 Jan 24 10:42 lon -> =-151d 52.155m
lrwxrwxrwx 1 16 Jan 24 10:42 name -> =Fish Creek Gage

This kind of thing would conventionally be stored in an rc file, but
having them available as files has some advantages. Each is exposed
and available without reading and parsing a file, for example.

Here's a shell function to read a symvar value:

symvar() {
test -L $1 || return
test -e $1 && return 1
value=$(readlink $1)
test "${value:0:1}" = = || return 1
value="${value:1}"
}

That returns the value in variable 'value', but in some cases printing
it may be preferred, e.g., in a utility program. The given name is
checked to see that it is a symlink and does not exist as a file, then
the value is read and checked to see that the first character is '='.

Writing a symvar is also simple to do, e.g.,

set_symvar() {
name=$1 && shift
ln -sf ="$*" $name
}

The -f force option will be needed if the symvar already exists, or that
can be left up to the caller via an option to the function.

I've tried different approaches for a symvar utility program, e.g.,

synopsis: symvar [-f] NAME [VALUE]
synopsis: symvar [-f] NAME[=VALUE] ...
synopsis: symvar [-f ]NAME[=VALUE] ...

where the latter allows several symvars to be queried or set at once,
but haven't settled on

In my opinion this is an unexploited and available aspect of the unix
filesystem, so why not put it to use?

Ken