Rough estimation of a C* table size

calc-table-size.py

#!/usr/bin/env python

import subprocess, shlex, re
from argparse import ArgumentParser


def tsize(type):
    sizes = {
        ("text", "varchar", "inet", "varint", "blob", "uuid", "timeuuid"): 16,
        ("bigint", "date", "decimal", "double", "timestamp", "time"): 8,
        ("boolean", "tinyint"): 1
    }

    for k, v in sizes.iteritems():
        if type in k:
            return v

    return 4


if __name__ == '__main__':
    parser = ArgumentParser()
    parser.add_argument('-t', '--table', required=True)
    parser.add_argument('-n', '--number', default=1, type=int)
    parser.add_argument('-p', '--partsize', default=1, type=int)
    args = parser.parse_args()

    p1 = subprocess.Popen(shlex.split("cqlsh -u ocs_superuser -p password -e \"desc smart.%s\"" % args.table),
                          stdout=subprocess.PIPE)
    p2 = subprocess.Popen(shlex.split("sed -n '/^CREATE/,/^)/{//!p}'"), stdin=p1.stdout, stdout=subprocess.PIPE)
    p1.stdout.close()

    desc = p2.communicate()[0].splitlines()
    colsizes = {}
    primary_keys = []
    clustering = []

    if "PRIMARY KEY" in desc[0]:
        primary_keys.append(desc[0].split()[0])
    elif "PRIMARY KEY" in desc[-1]:
        pk = re.sub(r'^.*KEY ', '', desc[-1])[1:-1]

        if pk[0] == '(':
            primary_keys += pk[pk.find("(") + 1: pk.find(")")].split(", ")
            clustering = pk[pk.find(")") + 3:].split(", ")
            primary_keys += clustering
        else:
            primary_keys = pk.split(", ")
            clustering = primary_keys[1:]

        desc = desc[:-1]

    for line in desc:
        linesplit = line.split()
        colsizes[linesplit[0]] = tsize(linesplit[1][:-1])

    print(colsizes)
    print(primary_keys)
    print(clustering)

    metadata_size = 8 * len(colsizes)
    pk_size = sum(colsizes[col] for col in primary_keys)
    ck_size = sum(colsizes[col] for col in clustering)
    reg_col_sizes = sum(colsizes.values()) - pk_size

    print(args.number * (pk_size + args.partsize * (reg_col_sizes + ck_size) + metadata_size))