Use sys.maxsize as upper bound for randrange

[ksindi]

It's possible to never exit the while loop if there are more than 1000 items that need to be bound. This PR mitigates it by using sys.maxsize, which is Python 2 and 3 compatible.

There is no real performance penalty by increasing the upper bound of random.randrange:

%timeit random.randrange(1, 1000)
# 773 ns ± 28.4 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
%timeit random.randrange(1, sys.maxsize)
# 822 ns ± 3.58 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

Reproducing error:

import string
import random

def generate_sql(items=None):
    sql_template = """
    SELECT
      col
    FROM
      table
    WHERE
      1
      {% if items %}
      AND table.col IN {{ items | inclause }}
      {% endif %}
    """

    j = JinjaSql()
    data = dict(items=items)
    sql, bind_params = j.prepare_query(sql_template, data)
    return sql, bind_params

def id_generator(size=6, chars=string.ascii_uppercase + string.digits):
    return ''.join(random.choice(chars) for _ in range(size))

# works
generate_sql(items=[id_generator() for _ in range(1000)])
# infinite loop!
generate_sql(items=[id_generator() for _ in range(1001)])

[ksindi]

cc @sripathikrishnan

[benrudolph]

ran into this too: https://github.com/hashedin/jinjasql/pull/17. though not sure we why wouldn't use hash, are we worried about performance?

[ksindi]

@benrudolph hashing is better. closing my PR in favor of yours.