Get data from a poorly formatted BLS table using Python

bls_data_get.py

import urllib2
import csv

FIRST_LINE = 11
LAST_LINE = 38

def get_level(l):
    for i, char in enumerate(l):
        if char != " ":
            break
    return i

def clean_line(l):
    l = l.replace("\r","")
    l = l.replace("\n","")
    l = l.split("    ")
    return [y.strip() for y in l if y!=""]

f = urllib2.urlopen("ftp://ftp.bls.gov/pub/suppl/empsit.tab1.txt")
lines = [line for line in f][FIRST_LINE:LAST_LINE]
levels = map(get_level, lines)
data_rows = map(clean_line, lines)
headings = [y[0] for y in data_rows]

d_order = dict(zip(headings, range(len(headings))))
d_level = dict(zip(headings, levels))

def one_up(heading):
    candidates = headings[:d_order[heading]]
    heading_level = d_level[heading]
    candidates.reverse()
    for c in candidates:
        if d_level[c] < heading_level:
            return c
    else:
        return None

def crumb_trail(heading):
    if one_up(heading) is None:
        return [heading]
    else:
        return crumb_trail(one_up(heading)) + [heading]

crumb_trails = map(crumb_trail, headings)
max_depth = max(map(len, crumb_trails))

g = open("bls.csv", "w")
header = ["level_%s" % i for i in range(max_depth)] + [
    "depth", "level", "normal_seasonal_movement", "estimated_over_month_change",
 "sa_adjusted_over_month_change"]

hier = dict(zip(set(levels), range(len(set(levels)))))
def get_hier(heading): return hier[d_level[heading]]

out = csv.writer(g)
out.writerow(header)
for trail, data_row in zip(crumb_trails, data_rows):
    industry = [None for i in range(max_depth)]
    indices = map(get_hier, trail)
    depth = max(indices)
    for name, i in zip(trail, indices):
        industry[i] = name
    out.writerow(industry + [depth] + data_row[:3])
g.close()