aws-error-utils 1.2.0

Error-handling functions for boto3/botocore

aws-error-utils

Making botocore.exceptions.ClientError easier to deal with

All AWS service exceptions are raised by boto3 as a botocore.exceptions.ClientError, with the contents of the exception indicating what kind of exception happened. This is not very pythonic, and the contents themselves are rather opaque, most being held in dicts rather than as properties. The functions in this package help dealing with that, to make your code less verbose and require less memorization of ClientError contents.

Installation

The package is on PyPI for pip-installing, but I tend to prefer just copying the aws_error_utils.py file into my projects; often then my only dependency is on boto3, which is usually somewhere in my environment anyway (e.g., in a Lambda function). But that's just me.

Usage

If you've got code like this:

s3 = boto3.client('s3')
try:
    s3.get_object(Bucket='my-bucket', Key='example')
except botocore.ClientError as error:
    if error.response['Error']['Code'] == 'NoSuchBucket':
        # error handling
    else:
        raise

you can replace it with:

s3 = boto3.client('s3')
try:
    s3.get_object(Bucket='my-bucket', Key='example')
except errors.NoSuchBucket:
    # error handling

If you have trouble remembering where all the contents in ClientError are, like these:

client_error.response['Error']['Code']
client_error.response['Error']['Message']
client_error.response['ResponseMetadata']['HTTPStatusCode']
client_error.operation_name

you can replace it with:

err_info = get_aws_error_info(client_error)

err_info.code
err_info.message
err_info.http_status_code
err_info.operation_name

If you're using errors or catch_aws_error(), you can skip the get_aws_error_info() step, because the fields are set directly on the ClientError object:

s3 = boto3.client('s3')
try:
    s3.get_object(Bucket='my-bucket', Key='example')
except errors.NoSuchBucket as error:
    error.code
    error.message
    error.http_status_code
    error.operation_name

errors

It's easiest to use the errors class if you don't have complex conditions to match. Using the error code as a field name in an except block will match that error code. Additionally, when you use this style, it sets the fields from AWSErrorInfo (see below) directly on the ClientError object. For example:

s3 = boto3.client('s3')
try:
    s3.get_object(Bucket='my-bucket', Key='example')
except errors.NoSuchBucket as error:
    print(error.message)

    # error handling

You can include multiple error codes in an except statement, though this is slower than combining them with a single catch_aws_error() call.

s3 = boto3.client('s3')
try:
    s3.get_object(Bucket='my-bucket', Key='example')
except (errors.NoSuchBucket, errors.NoSuchKey) as error:
    print(error.message)

    # error handling

You can only use this style for error codes that work as Python property names. For error codes like EC2's InvalidInstanceID.NotFound, you have to use catch_aws_error() (see below).

Unfortunately, you cannot get tab completion for error codes on the errors class, as a comprehensive list of error codes is not available as a Python package (botocore has a small number, but they are few and incomplete).

Note that the value of errors.NoSuchBucket is not an exception class representing the NoSuchBucket error, it is an alias for catch_aws_error('NoSuchBucket'). It can only be used in an except statement; it will raise RuntimeError otherwise. You also cannot instantiate the errors class.

catch_aws_error()

The function takes as input error code(s), and optionally operation name(s), to match against the current raised exception. If the exception matches, the except block is executed. If your error handling still needs the error object, you can still use an as expression, otherwise it can be omitted (just except catch_aws_error(...):). Additionally, catch_aws_error() sets the fields from AWSErrorInfo (see below) directly on the ClientError object.

s3 = boto3.client('s3')
try:
    s3.get_object(Bucket='my-bucket', Key='example')
except catch_aws_error('NoSuchBucket') as error:
    print(error.message)

    # error handling

You can provide error codes either as positional args, or as the code keyword argument with either as a single string or a list of strings.

catch_aws_error('NoSuchBucket')
catch_aws_error(code='NoSuchBucket')

catch_aws_error('NoSuchBucket', 'NoSuchKey')
catch_aws_error(code=['NoSuchBucket', 'NoSuchKey'])

If there are multiple API calls in the try block, and you want to match against specific ones, the operation_name keyword argument can help. Similar to the code keyword argument, the operation name(s) can be provided as either as a single string or a list of strings.

try:
    s3 = boto3.client('s3')
    s3.list_objects_v2(Bucket='bucket-1')
    s3.get_object(Bucket='bucket-2', Key='example')
except catch_aws_error('NoSuchBucket', operation_name='GetObject') as error:
    # This will be executed if the GetObject operation raises NoSuchBucket
    # but not if the ListObjects operation raises it

You must provide an error code. To match exclusively against operation name, use the aws_error_utils.ALL_CODES token. For completeness, there is also an ALL_OPERATIONS token.

try:
    s3 = boto3.client('s3')
    s3.list_objects_v2(Bucket='bucket-1')
    s3.get_object(Bucket='bucket-1', Key='example')
except catch_aws_error(ALL_CODES, operation_name='ListObjectsV2') as e:
    # This will execute for all ClientError exceptions raised by the ListObjectsV2 call

For more complex conditions, instead of providing error codes and operation names, you can provide a callable to evaluate the exception. Note that unlike error codes, you can only provide a single callable.

import re
def matcher(e):
    info = get_aws_error_info(e)
    return re.search('does not exist', info.message)

try:
    s3 = boto3.client('s3')
    s3.list_objects_v2(Bucket='bucket-1')
except catch_aws_error(matcher) as e:
    # This will be executed if e is a ClientError and matcher(e) returns True
    # Note the callable can assume the exception is a ClientError

get_aws_error_info()

This function takes a returns an AWSErrorInfo object, which is a collections.NamedTuple with the following fields:

• code
• message
• http_status_code
• operation_name
• response (the raw response dictionary)

If you're not modifying your except statements to use catch_aws_error(), this function can be useful instead of remembering exactly how this information is stored in the ClientError object.

If you're using catch_aws_error(), this function isn't necessary, because it sets these fields directly on the ClientError object.

aws_error_matches()

This is the matching logic behind catch_aws_error(). It takes a ClientError, with the rest of the arguments being error codes and operation names identical to catch_aws_error(), except that it does not support providing a callable.