Skip to main content

Lazy reading of file objects for efficient batch processing

Project description

lazyreader is a Python module for doing lazy reading of file objects.

The Python standard library lets you read a file a line-at-a-time, saving you from loading the entire file into memory. For example:

with open('large_file.txt') as f:
    for line in f:

lazyreader lets you do the same thing, but with an arbitrary delimiter, and for any object that presents a .read() method. For example:

from lazyreader import lazyread

with open('large_file.txt') as f:
    for doc in lazyread(f, delimiter=';'):

This is a snippet of code I spun out from the Wellcome Digital Platform. We have large XML and JSON files stored in S3 – sometimes multiple GBs – but each file is really a series of “documents”, separated by known delimiters. Downloading and parsing the entire file would be prohibitively expensive, but lazyreader allows us to hold just a single document in memory at a time.


lazyreader is available from PyPI:

$ pip install lazyreader


If we have a file stored locally, we can open it and split based on any choice of delimiter. For example, if we had a text file in which record were separated by commas:

with open('lots_of_records.txt') as f:
    for doc in lazyread(f, delimiter=','):

Another example: we have a file stored in Amazon S3, and we’d like to read it line-by-line. The boto3 API gives us a file object for reading from S3:

import boto3

client = boto3.client('s3')
s3_object = client.get_object(Bucket='example-bucket', Key='words.txt')
body = s3_object['Body']

for doc in lazyread(body, delimiter=b'\n'):

(This is the use case for which this code was originally written.)

One more example: we’re fetching an HTML page, and want to read lines separated by <br> in the underlying HTML. Like so:

import urllib.request

with urllib.request.urlopen('') as f:
    for doc in lazyread(f, delimiter=b'<br>'):

Advanced usage

lazyread() returns a generator, which you can wrap to build a pipeline of generators which do processing on the data.

First example: we have a file which contains a list of JSON objects, one per line. (This is the format of output files from elasticdump.) What the caller really needs is Python dictionaries, not JSON strings. We can wrap lazyread() like so:

import json

def lazyjson(f, delimiter=b'\n'):
    for doc in lazyread(f, delimiter=delimiter):

        # Ignore empty lines, e.g. the last line in a file
        if not doc.strip():

        yield json.loads(doc)

Another example: we want to parse a large XML file, but not load it all into memory at once. We can write the following wrapper:

from lxml import etree

def lazyxmlstrings(f, opening_tag, closing_tag):
    for doc in lazyread(f, delimiter=closing_tag):
        if opening_tag not in doc:

        # We want complete XML blocks, so look for the opening tag and
        # just return its contents
        block = doc.split(opening_tag)[-1]
        yield opening_tag + block

def lazyxml(f, opening_tag, closing_tag):
    for xml_string in lazyxmlstrings(f, opening_tag, closing_tag):
         yield etree.fromstring(xml_string)

We use both of these wrappers at Wellcome to do efficient processing of large files that are kept in Amazon S3.

Isn’t this a bit simple to be a module?

Maybe. There are recipes on Stack Overflow that do very similar, but I find it useful to have in a standalone module.

And it’s not completely trivial – at least, not for me. I made two mistakes when I first wrote this:

  • I was hard-coding the initial running string as

    running = b''

    That only works if your file object is returning bytestrings. If it’s returning Unicode strings, you get a TypeError (can’t concat bytes to str) when it first tries to read from the file. String types are important!

  • After I’d read another 1024 characters from the file, I checked for the delimiter like so:

    running += new_data
    if delimiter in running:
        curr, running = running.split(delimiter)
        yield curr + delimiter

    For my initial use case, individual documents were much bigger than 1024 characters, so the new data would never contain multiple delimiters. But with smaller documents, you might get multiple delimiters in one read, and then unpacking the result of .split() would throw a ValueError. So now the code correctly checks and handles the case where a single read includes more than one delimiter.

Now it’s encoded and tested in a module, I don’t have to worry about making the same mistakes again.



Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

lazyreader-1.0.1.tar.gz (4.5 kB view hashes)

Uploaded source

Built Distribution

lazyreader-1.0.1-py2-none-any.whl (7.2 kB view hashes)

Uploaded py2

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page