Attempt to build a complete AWS SDK in Perl
This project is attempting to build an entire AWS SDK from the information that is stored in other AWS SDKs. Other AWS SDKs have a "data-driven" approach, meaning that the definitions for the method calls are stored in a data structure describing input and output parameters.
The project is actually generating all of it's classes from botocore
Project info:
GitHub Actions status: 
Version on CPAN:
If you want to install and use Paws then just install it via cpan, cpanm, carton or the likes. If you want to contribute code: read on
cpanm Paws
echo "requires 'Paws';" >> cpanfile
carton install
carton exec my_script_that_uses_paws
If you want to develop a feature, or contribute code in some way, you need a development setup. This is done by cloning the repo into a local directory.
# Clone the repo. For what it's worth, you can clone from a fork too :)
git clone https://github.com/pplu/aws-sdk-perl.git
cd aws-sdk-perl
With carton you can install all the dependencies needed in a local environment, so you can play around with dependencies without affecting the system libraries. The cpanfile is used to track the dependencies needed.
It's possible that you need -dev libraries for compiling some of these modules. These packages are:
- In Debian/Ubuntu:
sudo apt-get install libxml2-dev libssl-dev
- In Red Hat/CentOS:
sudo yum install libxml2-devel openssl-devel
- In Mac OS X:
brew install openssl
If yo are using Mac OS X - El Capitan(10.11) you will probably need to force the link of the openssl header to /usr/local:
brew link openssl --force
And now tell carton to install the dependecies in a local lib
carton install
# drop into a shell so perl can always find the local libraries
carton exec $SHELL -l
Now we'll pull the paws-maintained fork of boto (so we can generate the SDK)
make pull-other-sdks
Now we're ready to code away! Happy hacking.
builder-lib: Contains classes that convert the botocore definitions into perl classes.
auto-lib: Contains the auto-generated classes. Changes to code in this directory will be overwritten, so only commit autogenerated code; never handwritten code (see "Generating API").
Changes that fix something in auto-lib are welcome, but cannot be applied directly, since they will be overwritten. Usually, they indicate some kind of general problem with other APIs, so the problem has to be fixed generically.
lib: Contains roles and classes that the auto-generated classes use to call the API, sign requests, handle responses, etc.
Note: This step is not necessary if you want to try out the SDK. We commit in "auto-lib" the classes generated by the definitions to which the "botocore" submodule points to. If you're not developing the SDK, go directly to the "Trying it out" step :)
Execute command make pull-other-sdks This will do a git pull of some official AWS sdks that are data-driven, and used to generate the SDK.
If documentation-1.json is not present in botocore/botocore/data/ for each service, which it will not when first using this project or when botocore possibly updates, then you need to first run the command to generate links to the AWS docs with:
make docu-links
This will allow the following two commands to function.
To generate the API for a given API call:
./gen_classes.pl --classes botocore/botocore/data/SERVICE/DATE/service-2.json
This will generate file(s) in auto-lib.
To generate all the APIs:
make gen-classes
The SDK is targeted at modern Perl versions. Since a new perl gets released every year, distributions perl tend to lag behind, so support for perl versions on any modern, widespread distribution is our target. Very old versions may work, but no intention to support them is made. You can always install a modern version of perl with perlbrew or plenv in a breeze. We're running the test cases on GitHub Actions for all "supported" perl versions. If you want to support a lower version, you can contribute back. Acceptance of patches for older versions of Perl won't mean that the compatibility will be maintained long-term, although it will be tried :).
Dependencies are versioned in a cpanfile. If you have carton, just execute 'carton install' in the sdk directory, and all dependencies will be pulled in automatically into a local library path. After that use 'carton exec ...' to execute your scripts.
If you add a dependency, just add it to the cpanfile file. There are three sections:
- the general section is for dependencies that are needed only in runtime
- the test section is for dependecies needed to run the test suite
- the develop section is for dependencies needed for developers
carton install installs all dependencies in all sections (after all, we're in developer mode here)
Packaging is managed with Dist::Zilla. To install Dist::Zilla and the necessary plugins, run:
make dist
After this, you will have a tar.gz suitable for uploading to CPAN or to your own mini-CPAN
The release process is as follows:
X.XX will be the version that you are currently working on (the one you want to release)
Y.YY will be the next version (which we start). This assumes origin as the main SDK source.
git checkout release/X.XX
git pull origin release/X.XX
Edit README.md to add any relevant contributions.
make gen-paws
Will copy the contents of the Contributions section of README.md into Paws.pm so that the same section appears in CPAN also.
Now edit the Changes file to add the date of release, adjusting the entry for version X.XX to
todays date.
Commit everything up till now and push
git push origin release/X.XX
Take a look at Travis pipelines to see if the branch you're going to release is green: https://travis-ci.org/pplu/aws-sdk-perl/branches
We don't want to ship Paws when it's failing it's tests.
If everything is OK:
git checkout master
git pull origin master
git merge release/X.XX
git push origin master
git tag release-X.XX
git push origin release-X.XX
make dist
this will generate Paws-X.XX.tar.gz, which is the artifact that is uploadable to CPAN
git checkout -b release/Y.YY
git push --set-upstream origin release/Y.YY
This creates the branch for working on the next release.
We bump the version number in the builder. Edit builder-lib/Paws/API/Builder/Paws.pm. Near line 12
we will find Paws version number. Replace X.XX for Y.YY
We add Y.YY to the Changes file. Commit the changes and push:
git push origin release/Y.YY
We're ready for developing Y.YY
After that: upload Paws-X.XX.tar.gz to CPAN an do a walk around issues and MRs that have been fixed / merged in the release, notifying that "Paws X.XX has hit CPAN with this issue fixed"
Each class for each API can be constructed in the following way:
Create a Perl script (myscript.pl)
#!/usr/bin/env perl
use Paws;
use Data::Printer;
my $iam = Paws->service('IAM');
my $summary = $iam->GetAccountSummary;
p $summary->SummaryMap;
also take a look at the CLI utility (see below) for fast testing)
See MetaCPAN https://metacpan.org/pod/Paws#AUTHENTICATION
or
perldoc Paws
Don't consider the SDK as "stable" code. There is a lot of experimenting going on. That said, people are using it in production, so changes to the way you call APIs, although not guaranteed, are not prone to change because they are autogenerated. Expect changes around the way you obtain service classes, transmit credentials, etc. Look at the TODO for expected changes to come.
As of 2015-02 I'm documenting breaking changes in the Changes file. API changes that break stuff will be documented there. Please read the Changes file before updating your git clone.
Although the code isn't considered stable yet, it works, and more than one person is using it already. I recommend you to using a cpanfile, and bundling Paws with carton bundle
Take a look at https://metacpan.org/pod/Paws#SUPPORTED-SERVICES
If a service is not supported, it will warn on construction with an explicit "non supported API" message. Basically all query and json services are supported. RestXML and RestJSON services are in the coming.
All services get auto-generated POD documentation. perldoc a file to take a look at the documentation.
Paws comes with a command-line utility to exercise the SDK. Just like Paws is the namespace for the SDK, "paws" (in /script) is the cli utility. It's quite rudimentary, but think of it as a quick way to try out calling services. Just call:
paws
to list all services. If a service isn't supported yet, it will die explicitly advising you that Paws doesn't support that service yet.
paws EC2 --region eu-west-1 DescribeInstances
Parameters needed in for DescribeInstances can be passed as a list of parameters (see https://metacpan.org/pod/ARGV::Struct) for complete documentation of how to pass parameters via command line.
paws EC2 --region eu-west-1 DescribeInstances Parameter1: ValueForParameter1 ListParameter: [ V1 V2 ] ComplexParam { Key1 Value1 Key2 Value 2 }
This code is distributed under the Apache v2 License
CAPSiDE (https://www.capside.com) for letting Paws be contributed in an open source model and giving me time to build and maintain it regularly.
ZipRecruiter (https://www.ziprecruiter.com/) for sponsoring development of Paws. Lots of work from ZipRecruiter has been done via Shadowcat Systems (https://shadow.cat/).
castaway for contributing to fixing documentation problems
- taking the reigns of Paws, become part of the core team that pushes it forward
- properly providing backlinks between related pages
- making TOCs render correctly on search.cpan.org
- generating helpful copy-paste ready scenarios in the synopsis of each method call
Luis Alberto Gimenez (@agimenez) for
- The git-fu cleaning up the "pull other sdks" code
- Credential Providers code
- Fixes for users that have no HOME env variable
- FileCaller to fully mock responses
Srinvas (@kidambisrinivas) for testing, bug reporting and fixing
juair10 for corrections and testing
CHORNY for CPAN and cpanfile packaging corrections
Iñigo Tejedor for service endpoint resolution based on rules
codehead for helping fix SQS Queue Maps
mbartold for helping fix SQS MessageBatch functionality
coreymayer for reporting bug in RestXmlCaller
arc (Aaron Crane) for documentation patches
dtikhonov for LWP Caller and bug reporting/fixing
vivus-ignis for DynamoDB bug reporting and test scripts for DynamoDB
karenetheridge for bug reporting, pull requests and help
ioanrogers for fixing unicode issues in tests
ilmari for fixing issues with timestamps in Date and X-Amz-Date headers, test fixes and 5.10 support fixes, documentation issue fixes for S3, CloudFront and Route53, help with number stringification
stevecaldwell77 for
- contributing support for temporary credentials in S3
- Fixing test suite failure scenarios
Ryan Olson (BeerBikesBBQ) for contributing documentation fixes
Roger Pettett for testing and contributing fixes for tests on MacOSX
Henri Yandell for help with licensing issues
Oriol Soriano (@ureesoriano) for contributions to API builders and better documentation generation
H. Daniel Cesario (@maneta) for devel setup instructions on RH and MacOSX
Glen van Ginkel for contributions to get S3 working
Javier Arellano for discovering Tagging bug
Ioan Rogers for contributing AssumeRoleWithSAML with ADFS auth example
Miquel Soriano for reporting a bug with DescribeAutoScalingGroups
Albert Bendicho (wiof) for contributing better retry logic
Brian Hartsock for better handling of XMLResponse exceptions
rpcme for reporting various bugs in the SDK
glenveegee for lots of work sorting out the S3 implementation
Grinzz
- many bugs, suggestions and fixes
- Installation speedup with Module::Builder::Tiny
Dakkar for solving issues with parameter passing
Arthur Axel fREW Schmidt for speeding up credential refreshing
PopeFelix for solving issues around S3 and MojoAsyncCaller
meis for (between others):
- contributing Paws::Credential::Explicit
- enabling unstable warnings to be silenced
sven-schubert for contributing fixes to RestXML services, working on fixing S3 to work correctly.
SeptamusNonovant for fixing paginators in non-callback mode
gadgetjunkie for contributing the ECS credential provider
mla for contributing a fix to correct dependencies
autarch for correcting signature generation for a bunch of services
piratefinn for linking calls to documentation AWS URLs
slobo for fixing S3 behaviour
bork1n for fixes to MojoAsynCaller
atoomic for:
- tweaking CPAN packaging
- improving paws CLI
leonerd for (between others)
- documenting retry logic
- fixing retry sleep of MojoAsyncCaller
campus-explorer for contributing to test suite
byterock for:
- testing and fixing PinPoint
- standing up as comaint, and releasing 0.43
- improving S3 support
torrentale for fixing QueryCaller to correctly signal empty arrays
Jess Robinson and shadowcat.co.uk for:
- doing lots of comaint work
- working hard on new features
shogo82148 for migrating our Travis pipelines to GitHub Actions (and improving them)
aeruder for contributing
- Fixing DynamoDB retry fixes
- Completing speedups and benchmarking code
- Substituting Config::INI for Config::AWS
- Parrallelizing and fixing generation inconsistencies of the SDK
dheffx for making ContainerProfile credential provider more robust
0leksii for building support for Instance Metadata Service v2 (IMDSv2)