How do I perform binary search on a text file to search a keyword in python?

Here's an interesting way to do it with Python's built-in bisect module.

import bisect
import os


class Query(object):

    def __init__(self, query, index=5):
        self.query = query
        self.index = index

    def __lt__(self, comparable):
        return self.query < comparable[self.index:]


class FileSearcher(object):

    def __init__(self, file_pointer, record_size=35):
        self.file_pointer = file_pointer
        self.file_pointer.seek(0, os.SEEK_END)
        self.record_size = record_size + len(os.linesep)
        self.num_bytes = self.file_pointer.tell()
        self.file_size = (self.num_bytes // self.record_size)

    def __len__(self):
        return self.file_size

    def __getitem__(self, item):
        self.file_pointer.seek(item * self.record_size)
        return self.file_pointer.read(self.record_size)


if __name__ == '__main__':
    with open('data.dat') as file_to_search:
        query = raw_input('Query: ')
        wrapped_query = Query(query)

        searchable_file = FileSearcher(file_to_search)
        print "Located @ line: ", bisect.bisect(searchable_file, wrapped_query)

Do you need do do a binary search? If not, try converting your flatfile into a cdb (constant database). This will give you very speedy hash lookups to find the index for a given word:

import cdb

# convert the corpus file to a constant database one time
db = cdb.cdbmake('corpus.db', 'corpus.db_temp')
for line in open('largecorpus.txt', 'r'):
    index, word = line.split()
    db.add(word, index)
db.finish()

In a separate script, run queries against it:

import cdb
db = cdb.init('corpus.db')
db.get('chaos')
12345

If you need to find a single keyword in a file:

line_with_keyword = next((line for line in open('file') if keyword in line),None)
if line_with_keyword is not None: 
   print line_with_keyword # found

To find multiple keywords you could use set() as @kriegar suggested:

def extract_keyword(line):
    return line[5:35] # assuming keyword starts on 6 position and has length 30

with open('file') as f:
    keywords = set(extract_keyword(line) for line in f) # O(n) creation
    if keyword in keywords: # O(1) search
       print(keyword)

You could use dict() above instead of set() to preserve index information.

Here's how you could do a binary search on a text file:

import bisect

lines = open('file').readlines() # O(n) list creation
keywords = map(extract_keyword, lines) 
i = bisect.bisect_left(keywords, keyword) # O(log(n)) search
if keyword == keywords[i]:
   print(lines[i]) # found

There is no advantage compared to the set() variant.

Note: all variants except the first one load the whole file in memory. FileSearcher() suggested by @Mahmoud Abdelkader don't require to load the whole file in memory.

I wrote a simple Python 3.6+ package that can do this. (See its github page for more information!)

Installation: pip install binary_file_search

Example file:

1,one
2,two_a
2,two_b
3,three

Usage:

from binary_file_search.BinaryFileSearch import BinaryFileSearch
with BinaryFileSearch('example.file', sep=',', string_mode=False) as bfs:
    # assert bfs.is_file_sorted()  # test if the file is sorted.
    print(bfs.search(2))

Result: [[2, 'two_a'], [2, 'two_b']]

It is quite possible, with a slight loss of efficiency to perform a binary search on a sorted text file with records of unknown length, by repeatedly bisecting the range, and reading forward past the line terminator. Here's what I do to look for look thru a csv file with 2 header lines for a numeric in the first field. Give it an open file, and the first field to look for. It should be fairly easy to modify this for your problem. A match on the very first line at offset zero will fail, so this may need to be special-cased. In my circumstance, the first 2 lines are headers, and are skipped.

Please excuse my lack of polished python below. I use this function, and a similar one, to perform GeoCity Lite latitude and longitude calculations directly from the CSV files distributed by Maxmind.

Hope this helps

========================================

# See if the input loc is in file 
def look1(f,loc):
# Compute filesize of open file sent to us
hi = os.fstat(f.fileno()).st_size
lo=0
lookfor=int(loc)
# print "looking for: ",lookfor
while hi-lo > 1:
    # Find midpoint and seek to it
    loc = int((hi+lo)/2)
    # print " hi = ",hi," lo = ",lo
    # print "seek to: ",loc
    f.seek(loc)
    # Skip to beginning of line
    while f.read(1) != '\n':
        pass
    # Now skip past lines that are headers
    while 1:
        # read line
        line = f.readline()
        # print "read_line: ", line
        # Crude csv parsing, remove quotes, and split on ,
        row=line.replace('"',"")
        row=row.split(',')
        # Make sure 1st fields is numeric
        if row[0].isdigit():
            break
    s=int(row[0])
    if lookfor < s:
        # Split into lower half
        hi=loc
        continue
    if lookfor > s:
        # Split into higher half
        lo=loc
        continue
    return row  # Found
# If not found
return False

Consider using a set instead of a binary search for finding a keyword in your file.

Set:

O(n) to create, O(1) to find, O(1) to insert/delete

If your input file is separated by a space then:

f = open('file')
keywords = set( (line.strip().split(" ")[1] for line in f.readlines()) )
f.close()    

my_word in keywords
<returns True or False>

Dictionary:

f = open('file')
keywords = dict( [ (pair[1],pair[0]) for pair in  [line.strip().split(" ") for line in f.readlines()] ] ) 
f.close()

keywords[my_word]
<returns index of my_word>

Binary Search is:

O(n log n) create, O(log n) lookup

edit: for your case of 5 characters and 30 characters you can just use string slicing

f = open('file')
keywords = set( (line[5:-1] for line in f.readlines()) )
f.close()

myword_ in keywords

or 

f = open('file')
keywords = dict( [(line[5:-1],line[:5]) for line in f.readlines()] )
f.close()

keywords[my_word]