awk implementation of geohash

requirements: awk > 4.1 (support for the --bignum flag)

Inspired by a blog post.

geohash accepts all arguments permitted by awk as well as the following:

• lat, lon (columns number with latitudes and longitudes)
• precision (number of characters in the geohash)
• geohash (column number with a geohash)

geohash accepts a text file with columns of data. To encode coordinates, specify lat and lon columns (precision is optional). To decode geohash to coordinates specify geohash column. The output columns are added to the original data and printed in stdout. lat and lon are printed with 16 digits precision.

Examples

% echo "19.28924560546875 33.5467529296875 berrybush" | geohash lat=1 lon=2 geohash=3 precision=7
19.28924560546875 33.5467529296875 berrybush servers 64.6483612060546875 -147.0017051696777344
%
% echo "othercolumn,-47.17529296875,68.35693359375" | geohash -F, lat=3 lon=2
othercolumn,-47.17529296875,68.35693359375,funkys000000

Note that invalid characters in geohash (invalid symbols in coordinates) are treated as zero.

% echo dog | geohash geohash=1 | geohash lat=2 lon=3 precision=3
dog 4.2187500000000000 -85.7812500000000000 d0g
%
% echo "0 5\ndog 5" | geohash lat=1 lon=2
0   5 s0581b0bh2n0
dog 5 s0581b0bh2n0

Benchmarking

Comparing to an equivalent python script the awk version works ~3 times slower. I attribute it to less efficient bitwise operations in awk.

% cat > script.py
import geohash

with open('test.csv', 'r') as f:
    for line in f:
        print(geohash.encode\
              (*[float(x) for x in line.rstrip().split(',')]))
%
% wc -l test.csv
1000000 test.csv
% /bin/time -v geohash -F, lat=1 lon=2 test.csv > /dev/null
    ...
    Elapsed (wall clock) time (h:mm:ss or m:ss): 0:12.06
    ...
    Maximum resident set size (kbytes): 4996
    ...
% /bin/time -v python3 script.py > /dev/null
    ...
    Elapsed (wall clock) time (h:mm:ss or m:ss): 0:04.00
    ...
    Maximum resident set size (kbytes): 8580
    ...