High-precision scientific notation

Discussions related to the extensions to the language available in LBB
guest
Site Admin
Posts: 192
Joined: Tue Apr 03, 2018 1:34 pm

High-precision scientific notation

Post by guest »

(copied from the old forum)

LB 4 (including 4.5.1) performs floating-point arithmetic internally with a precision of approximately 15 to 16 significant figures, which should be good enough for most applications. But, strangely, if the calculation returns a value that must be represented in scientific (exponential) notation - because it is too large or too small for fixed-point format - LB will print the result only to a precision of 8 significant figures!

For example suppose you want to calculate 123456789^-20; LB knows what the answer is to an accuracy of at least 15 digits but the best you can print (easily) is:

Code: Select all

0.14780886e-161
This limitation arises because, uniquely in my experience, the LB 4 implementation of the USING() function does not give you the opportunity of specifying that scientific notation should be used. This omission is fixed in LBB:

Code: Select all

    print using ("###.###################^^^^", 123456789^-20)
which gives:

Code: Select all

1.4780885631555544717E-162
(all the digits are significant).

Here's a program for printing all the powers from +42 to -42 (only a fraction of the range that LBB is capable of):

Code: Select all

    a = 123456789
    for n = 42 to -42 step -1
        b = a^n
        print n;tab (10); using ("###.###################^^^^",b)
    next n
    end 
Richard.