I've tried many different ways of evaluating the mess below as a float. I think I'm almost there with the use of bc but it's still not evaluating.
s=$((($s+((${flowx[$(($a+1))]}-(${flowx[$a]}))*(${flowy[$(($a+1))]}+${flowy[$a]}))/2) | bc))
Any input?
Don't use $(( )) at all.
For instance, taking a relevant part of your calculation:
s=$(bc <<<"$s + ( ( ${flowx[a+1]} - ${flowx[a]} ) * ( ${flowy[a+1]} + ${flowy[a]} ) ) / 2")
As long as flowx and flowy are standard integer-indexed arrays rather than associative arrays, you don't even need to use $(( )) when indexing into them (or $ operators inside those indices), as the index of a nonassociative array in bash is a math context by default.
Or, more readably than all these nested parens, use dc (here, configured with 10 digits of precision):
s=$(dc <<EOF
10 k
${flowx[a+1]}
${flowx[a]}
-
${flowy[a+1]}
${flowy[a]}
+
*
$s +
2 /
p
EOF
)
See? Much more readable.
<<< is a bashism -- if you're using /bin/sh rather than bash, there's potential for an error there. Beyond that, change the bc <<< to an echo and hand-check the expansion -- this is off-the-cuff rather than something that could be tested, since you didn't provide samples of your arrays' values &c. sufficient to allow testing of any proposed solution.
flowax to flowx in the bc line... ;)<<<, exactly as given? If you wrote, say, <"..." rather than <<<"...", that would do what you described. (Even if that guess is incorrect, it'd be helpful if you provided a copy-and-paste of the invocation).You can avoid all quotes and such things if you instead of the hereline (as @CharlesDuffy in the otherwise great answer recommends) will use the heredoc as like:
flowx=(0.124852 -0.0156593 -0.0932662 -0.0464323 0.0305706 0.00833429 0.0245359 0.0292034 -0.0564935)
flowy=(0.197532 0.120311 0.0864692 -0.0071995 0.097294 0.0624036 0.0825287 0 0.0340206)
a=1
s=0.1
res=$(bc -l <<EOF
$s + ( ( ${flowx[a+1]} - ${flowx[a]} ) * ( ${flowy[a+1]} + ${flowy[a]} ) ) / 2
EOF
)
echo "$res"
prints
.09197621484831000000
If ksh93 is installed on your system, I suggest you use it instead of bash for this task. It has built-in support for 64 bit IEEE floating point.
bc, which... well, yeah. Usebc, don't use$(( )). Since you already know thatbcexists, and that it can do floating-point math, how is this a question?dcgives you true arbitrary-precision math rather than the compromises inherent in IEEE floating-point. It's actually more accurate (if asked to be) than what awk/perl/python/ruby/etc. do out-of-the-box with floats.