an arbitrary precision maths library written entirely in Bourne shell

In late 2001, I wrote the skeleton of a pure /bin/sh implementation of various mathematical operations (doesn't even use test or echo). It includes the boolean operators and conversion between binary and decimal, plus a few helper functions. The code is notionally arbitrary-precision, but will in fact be limited by how long a particular Bourne shell implementation allows variables to be.

A sample script calculates a broadcast address given IP address and netmask, and is merely proof-of-concept code to disprove Randal L. Schwartz's assertion, on the SAGE mailing list, that it couldn't be done. He had written:

There are no solutions in sh (other than brute force). Sh has insufficient computing ability - it'll have to call some other program (and I bet a lot of the answers will probably try using bc or expr).

Of course, Clarke's First Law applies here.

Coming back to it a year later, I realized that the code doesn't understand negative numbers. I'm not entirely sure how to rectify that; While the code does have all the boolean operators, it is arbitrary precision, and so I don't think normal digital-computer like operations will work. It's been over a decade since I studied that sort of thing. And, do I really care?

webmaster @ jhnc.org / 2003-02-05 (updated 2018-06-26)