Skip to main content

An S3 signed URL generator, faster than boto3 and aiobotocore's, for generating lots of signed URLs quickly.

Project description

fast_s3_url

An S3 URL generator that's (much) faster than boto3/aiobotocore's, while being compliant with the s3v4 signed URL format. How much faster? See below.

Largely based upon: https://github.com/WeTransfer/wt_s3_signer.

Important notice

This library, much like the one it's based upon, skips several steps that boto3 and aiobotocore do. As a result, this library should be used one very specific use case only -- when you need to generate thousands of signed URLs fast, and most importantly, when you control the object keys themselves.

For any other use case scenario, please use boto3 or aiobotocore instead.

Installation

After making sure this is for you, with your favourite package manager (e.g., pip):

pip install fast-s3-url

Recommended reading

Usage

Note: This library does not depend on boto3 or aiobotocore explicitly. If you want to use it without them, you can create an instance directly (see below).

With boto3

Use FastS3UrlSigner.from_boto3_client to create an instance from an existing (and properly configured) client.

client = boto3.client("s3") 

signer = FastS3UrlSigner.from_boto3_client(client, bucket="test")

object_keys = ["cat.jpg", ...]

signed_urls = signer.generate_presigned_get_object_urls(object_keys, expires_in=3600)

With aiobotocore

Use FastS3UrlSigner.from_aiobotocore_client to create an instance from an existing (and properly configured) client. Note that unlike the boto3 variant, this method needs to be awaited.

session = get_session()

async with session.create_client("s3") as client:
    signer = await FastS3UrlSigner.from_aiobotocore_client(client, "test")

    object_keys = ["cat.jpg", ...]

    signed_urls = signer.generate_presigned_get_object_urls(object_keys, expires_in=3600)

Manually creating an object

The following example assumes you have a MinIO instance running on http://localhost:9000 with default settings.

Note that you shouldn't hardcode credentials, this is merely for illustration purposes.

from fast_s3_signer import FastS3UrlSigner, Credentials

signer = FastS3UrlSigner(
    bucket_endpoint_url="http://localhost:9000/my-bucket/",
    aws_region="us-east-1",
    credentials=Credentials(
        access_key="minioadmin",
        secret_key="minioadmin",
    ),
)

:warning: Note on instances created from boto3 and aiobotocore

The boto3 and aiobotocore clients may, when they are configured with an STS Security Token, refresh their credentials when generating signed URLs. When a FastS3UrlSigner is created with these clients, their credentials are refreshed so that the signer has access to the current credentials, but it won't refresh them after it's created. The signer doesn't even keep a reference to the client you use to create it!

This means that if your aiobotocore/boto3 clients happen to have expiring tokens in them, URLs created by FastS3UrlSigner may suddently stop working. Therefore, the best way to use a FastS3UrlSigner created from a client is to create it in a function-local scope, use it, and then have it be destroyed by GC when you no longer need it.

If you create your signer instances with non-expiring credentials, you can safely disregard this warning.

Local development

This project uses Poetry for dependency management. Make sure you have it installed.

You can then create a Virtual Environment with it (any Python version >= 3.8 will do), and set up a shell. See Poetry's documentation for tutorials on how to do this.

Running tests

If you have make installed, the following targets are set up for you:

  • unit-test: Runs unit tests. Doesn't require any S3 instance.
  • integration-test: Runs integration tests. Requires an S3 instance running on http://localhost:9000. You can use MinIO to create one.

Note: These targets assume you are running a shell with a Poetry Virtual Environment set up. They do not spawn a shell for running the commands.

Benchmarks

The following results are from Python 3.11 on Fedora 37 running on a Lenovo ThinkPad L390, with an Intel Core i7 8565U and 16 GB RAM. Turbo Boost was disabled because it quickly causes throttling and it can skew results.

As shown below, depending on the number of URLs to generate, this library can be 6 times faster than the official libraries when generating a single URL, and over 30 times faster when generating 100 or more URLs.

You can run the following to run the benchmarks with pytest-benchmark:

make benchmark
---------------------------------------------------------------- benchmark 'signed_url_generation object_key_count=1': 2 tests ----------------------------------------------------------------
Name (time in us)          Min                 Max                Mean            StdDev              Median                IQR            Outliers  OPS (Kops/s)            Rounds  Iterations
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_custom[1]         55.5300 (1.0)      103.7830 (1.0)       57.4275 (1.0)      2.1685 (1.0)       56.8480 (1.0)       0.6910 (1.0)     1187;1741       17.4133 (1.0)       16791           1
test_boto3[1]         355.7460 (6.41)     492.3180 (4.74)     367.2364 (6.39)     6.9269 (3.19)     364.7620 (6.42)     10.6518 (15.41)      659;13        2.7230 (0.16)       2621           1
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

------------------------------------------------------------ benchmark 'signed_url_generation object_key_count=1024': 2 tests -----------------------------------------------------------
Name (time in ms)          Min                 Max                Mean            StdDev              Median               IQR            Outliers      OPS            Rounds  Iterations
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_custom[1024]      11.3772 (1.0)       11.5678 (1.0)       11.4341 (1.0)      0.0283 (1.0)       11.4352 (1.0)      0.0308 (1.0)          19;2  87.4577 (1.0)          88           1
test_boto3[1024]      427.2083 (37.55)    429.6503 (37.14)    427.9307 (37.43)    0.9811 (34.63)    427.6766 (37.40)    0.7958 (25.81)         1;1   2.3368 (0.03)          5           1
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

---------------------------------------------------------- benchmark 'signed_url_generation object_key_count=128': 2 tests -----------------------------------------------------------
Name (time in ms)         Min                Max               Mean            StdDev             Median               IQR            Outliers       OPS            Rounds  Iterations
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_custom[128]       1.4635 (1.0)       1.6152 (1.0)       1.4895 (1.0)      0.0110 (1.0)       1.4889 (1.0)      0.0114 (1.0)        131;11  671.3848 (1.0)         665           1
test_boto3[128]       53.6779 (36.68)    55.7975 (34.54)    54.3269 (36.47)    0.6903 (63.01)    53.9923 (36.26)    0.9660 (85.04)         4;0   18.4071 (0.03)         19           1
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

------------------------------------------------------------ benchmark 'signed_url_generation object_key_count=512': 2 tests -------------------------------------------------------------
Name (time in ms)          Min                 Max                Mean            StdDev              Median               IQR            Outliers       OPS            Rounds  Iterations
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_custom[512]        5.7369 (1.0)        5.8751 (1.0)        5.8002 (1.0)      0.0266 (1.0)        5.8028 (1.0)      0.0361 (1.0)          53;1  172.4078 (1.0)         172           1
test_boto3[512]       215.7762 (37.61)    219.7181 (37.40)    217.4519 (37.49)    1.6630 (62.51)    217.8732 (37.55)    2.6384 (73.16)         2;0    4.5987 (0.03)          5           1
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

------------------------------------------------------------------- benchmark 'signed_url_generation object_key_count=8': 2 tests --------------------------------------------------------------------
Name (time in us)            Min                   Max                  Mean             StdDev                Median                IQR            Outliers         OPS            Rounds  Iterations
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_custom[8]          133.2900 (1.0)        189.1850 (1.0)        137.3525 (1.0)       3.2919 (1.0)        136.2515 (1.0)       1.6770 (1.0)     1137;1124  7,280.5388 (1.0)        7182           1
test_boto3[8]         2,829.3420 (21.23)    4,067.5490 (21.50)    2,869.7947 (20.89)    82.0675 (24.93)    2,854.3680 (20.95)    13.6910 (8.16)        16;43    348.4570 (0.05)        349           1
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

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

fast_s3_url-0.2.0.tar.gz (11.4 kB view details)

Uploaded Source

Built Distribution

fast_s3_url-0.2.0-py3-none-any.whl (9.2 kB view details)

Uploaded Python 3

File details

Details for the file fast_s3_url-0.2.0.tar.gz.

File metadata

  • Download URL: fast_s3_url-0.2.0.tar.gz
  • Upload date:
  • Size: 11.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.2.2 CPython/3.11.0 Linux/6.0.11-300.fc37.x86_64

File hashes

Hashes for fast_s3_url-0.2.0.tar.gz
Algorithm Hash digest
SHA256 6174f408aab081eedd429d7b0b69f5210b55c9d7a6f10aac8c357ec97549f5de
MD5 948a2a7e04406ff09b28d22baca3f7a8
BLAKE2b-256 14d61aec2eb99a59a20add1fc9670b94dc3cafe86b6c199802eac2eb2441920d

See more details on using hashes here.

File details

Details for the file fast_s3_url-0.2.0-py3-none-any.whl.

File metadata

  • Download URL: fast_s3_url-0.2.0-py3-none-any.whl
  • Upload date:
  • Size: 9.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.2.2 CPython/3.11.0 Linux/6.0.11-300.fc37.x86_64

File hashes

Hashes for fast_s3_url-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 b5b993844b7cd7a3e981b98b1fbd61455acbd6a091cf144ef3939e241397ff9a
MD5 dca6762f641157cd18501ce5a162988a
BLAKE2b-256 e69ffb34ddf892efb587f2825c5c067d6ae655f77af122f7211528dd2a8c0843

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page