simplecep 0.1.1

Fetch CEP addresses consistently using Correios API, third-party APIs as fallbacks and cache the results.

simpleCEP

Reliably resolve Brazilian CEP addresses using multiple APIs.

Why

ECT (Correios Company) doesn't open their CEP (Brazilian Zip Codes) data, requiring systems to fetch CEP data individually through their API or buying the full database from their website.

Correios API (SIGEPWeb) is known for its outages, which impacts systems which relies on resolving CEPs as addresses, as e-commerces.

This library addresses the problem using alternative CEP data APIs as fallbacks when Correios API doesn't respond in time and caching all received results.

How it works

The library uses the configured CEP data providers to fetch the CEP address. Each installed provider is used until the data is successfully fetched.

It will try to get the CEP address using Correios official API and if it fails it will use third-party API as fallbacks.

By default, it uses the following CEP APIs:

• https://www.republicavirtual.com.br
• https://viacep.com.br/

Both APIs are queried concurrently using threads, and the first received result is used.

After fetched, the retrieved address data is cached so any address-retrieval attempt for the same CEP is resolved immediately and won't use external providers.

The used providers, their order, and the cache mechanism can be fully customized.

If the CEP data isn't cached and all providers fail to retrieve it, the CouldNotResolveCepError exception is raised.

Usage

Installation

Install from PyPI

pip install simplecep

Using the default config

The package provides a ready-to-use resolve_cep function which will try to, in order:

1. Get the CEP address from the internal cache
2. Get the CEP address from the official Correios API
3. Get the CEP address, in parallel, from third-party CEP APIs:
   • https://www.republicavirtual.com.br
   • https://viacep.com.br/

It will stop on success, and the subsequent steps won't be executed. By default, it will wait 2 seconds for each provider.

Example:

>> from simplecep import resolve_cep
>> address = resolve_cep('59615350')

>> address
<CEPAddress 59615-350>

>> address.to_dict()
{
    'cep': '59615-350',
    'street': 'Rua João Simão do Nascimento',
    'district': 'Santa Delmira',
    'city': 'Mossoró',
    'state': 'RN'
}

The CEPAddress class provides the address fields in both English and Brazilian Portuguese names:

>> address
<CEPAddress 59615-350>

>> address.street
'Rua João Simão do Nascimento'

>> address.rua
'Rua João Simão do Nascimento'

>> address.city == address.cidade
True

>> address.to_dict(br_names=True)
{
    'cep': '59615-350',
    'rua': 'Rua João Simão do Nascimento',
    'bairro': 'Santa Delmira',
    'cidade': 'Mossoró',
    'estado': 'RN'
}

Customization

If you need to use the providers in a different order, a custom provider, a different cache mechanism or change the timeout, a new resolver instance can be created using the CEPResolver class.

Here's how the default resolve_cep function is created:

from simplecep import (
    CEPResolver,
    correios_sigep_cep_provider,
    republicavirtual_cep_provider,
    viacep_cep_provider
)

resolve_cep = CEPResolver(
    providers=[
        # this resolver will first try to use the Correios API,
        {correios_sigep_cep_provider},
        # if it fails it will fall back to the RepublicaVirtual and ViaCEP APIs
        # it will spawn two threads and will query both APIs at the same time
        # the result from the first one to reply will be used
        {republicavirtual_cep_provider, viacep_cep_provider},
    ],
    # by default it will wait for each provider for 1 second
    timeout=2,
    # all fetched CEP data will be stored in the
    # cache object which is a dict but can be any dict-like object
    cache={}
)

Providers Order

The providers argument should be a list of sets specifying which providers should be used, and the order they should be called.

When called, the resolver iterate over the list calling all providers in each set concurrently using threads.

When the CEP is successfully resolved by a provider, the iteration stops and the following provider sets are not executed.

Here's an example of providers param which will query all APIs concurrently and use the first received result:

brute_resolver = CEPResolver(
    providers=[
        {
            correios_sigep_cep_provider,
            republicavirtual_cep_provider,
            viacep_cep_provider
        }
    ]
)

Custom providers

Custom providers can be used to fetch CEP data from different sources.

The provider should be a callable which accepts cep and timeout as parameters and returns a CEPAddress object when a CEP is found and None when the CEP doesn't exist:

def my_custom_provder(cep: str, timeout: float) -> Optional[CEPAdress]:
    pass # my implementation here

Custom providers should raise the CEPProviderUnavailableError error when they fail to contact their data source or if the data source doesn't reply before the specified timeout interval.

Provider-auxiliary functions are available in the simplecep.providers.commons module.

Custom cache

By default, a dict object is used as cache, it's good enough to avoid querying for the same CEP twice in the same process, but it has limitations as:

• Different processes can't share the cache (i.e. web-workers)
• The cache is destroyed when the process is killed

So a custom object can be used as cache, it needs to be a dict-like object (i.e. implements __getitem__ and __setitem__).

A good example is the redis-dict project which creates a dict-like object which stores the values on Redis: https://pypi.org/project/redis-dict/

Testing

Tests are found in the tests folder. Install the requirements in the dev_requirements.txt file and run make test to run the tests.

License

See License.

History

0.1.1 (2021-03-29)

• CEPAddress now supports Brazilian address fields names (estado, cidade, bairro, rua)
• Update capture_real_responses.py test-script to not double-request providers
• Clean RepublicVirtual provider "address" field

0.1.0 (2021-03-12)

• First release on PyPI