Skip to content

Python package supporting ODS OpenDataStandards (OED / ORD)

Notifications You must be signed in to change notification settings

OasisLMF/ODS_Tools

Repository files navigation

ODS Tools

Overview

ODS Tools is a Python package designed to support users of the Oasis Loss Modelling Framework (Oasis LMF). This package includes a range of tools for working with Oasis data files, including loading, conversion and validation.

The package is based on a release of: in accordance with the ODS_OpenExposureData.

Installation

ODS_Tools can be installed via pip by running the following command:

pip install ods-tools

If using the transform command, use instead

pip install ods-tools[extra]

which will install some optional packages needed for it.

command line interface

ODS tools provide command line interface to quickly convert oed files:

example :

ods_tools convert --location path_to_location_file --path output folder

see ods_tools convert --help for more option

Usage

loading exposure data

in order to load oed file we use the concept of source. A source will define how to retrieve the oed data. For the moment we only support files but other type of source such as DataBase could be envisaged. The loading itself support several format such as parquet, csv and all pandas read_csv supported compression The path to the file can be absolute relative or even an url

config example:

config = {
    'location': 'SourceLocOEDPiWind.csv', # csv file
    'account': 'SourceAccOEDPiWind.parquet', # parquet file
    'ri_info': {
        'cur_version_name': 'orig', # passing args to the reader function
        'sources': {
            'orig': {
                'source_type': 'filepath',
                'filepath': 'SourceReinsInfoOEDPiWind.csv',
                'read_param': {
                    'usecols':[
                        'ReinsNumber', 'ReinsLayerNumber', 'ReinsName', 'ReinsPeril',
                        'ReinsInceptionDate', 'ReinsExpiryDate', 'CededPercent', 'RiskLimit',
                        'RiskAttachment', 'OccLimit', 'OccAttachment', 'PlacedPercent',
                        'ReinsCurrency', 'InuringPriority', 'ReinsType', 'RiskLevel', 'OEDVersion'
                    ]
                }
            }
        }
    },
    'ri_scope': 'https://raw.githubusercontent.com/OasisLMF/OasisPiWind/main/tests/inputs/SourceReinsScopeOEDPiWind.csv', # url
}

Access Oed File as DataFrame

Once the config is done you can create your OedExposure Object and access the Dataframe representation of the different sources. Data Type in the DataFrame will correspond to the type

import ods_tools
oed_exposure = ods_tools.oed.OedExposure(**config)
location = oed_exposure.location.dataframe
account = oed_exposure.account.dataframe

Saving Change to the oed DataFrame

You can modify the DataFrame and save it as a new version

oed_exposure.location.save(version_name='modified version',
                            source='path_to_save_the_file')

you may also save the exposure itself this will save the current dataframe to the specified directory_path. if you specify version_name, oed files will be saved as f'{version_name}_{OED_NAME}' + compression (ex: version_2_location.csv) if the version_name is an empty string, oed files will be saved as just f'{OED_NAME}' + compression (ex: location.parquet) if version_name is None, oed files will take the same name as the current source if it is a filepath or f'{OED_NAME}' + compression otherwise (ex: SourceLocOEDPiWind.csv)

compression let you specify the file extension (csv, parquet, zip, gzip, bz2, zstd)

if save_config is True the exposure config will also be saved in the directory

oed_exposure.save(directory_path, version_name, compression, save_config)

OED Validation

Validity of oed files can be checked at loading time with the argument check_oed

oed_exposure = ods_tools.oed.OedExposure(check_oed=True, validation_config=validation_config, **config)

validation_config is a list of all check that you want to perform, if one oed source fail a check depending on validation_config 4 different action can be performed 'raise', 'log', 'ignore', 'return'.

  • 'raise' will raise an OdsException
  • 'log' will log the issue with a info level
  • 'ignore' will ignore the issue
  • 'return' will return the check issue in a list in order for the developer to perform its own treatment. In that case the check method need to be called instead of relying on the constructor
oed_exposure = ods_tools.oed.OedExposure(check_oed=False**config)
invalid_data = oed_exposure.check(custom_validation_config)

the curent default validation under ods_tools.oed.common DEFAULT_VALIDATION_CONFIG contains

VALIDATOR_ON_ERROR_ACTION = {'raise', 'log', 'ignore', 'return'}
DEFAULT_VALIDATION_CONFIG = [
    {'name': 'required_fields', 'on_error': 'raise'},
    {'name': 'unknown_column', 'on_error': 'log'},
    {'name': 'valid_values', 'on_error': 'raise'},
    {'name': 'perils', 'on_error': 'raise'},
    {'name': 'occupancy_code', 'on_error': 'raise'},
    {'name': 'construction_code', 'on_error': 'raise'},
    {'name': 'country_and_area_code', 'on_error': 'raise'},
]

An OdsException is raised with a message indicating which file is invalid and why.

Currency Conversion Support

Exposure Data handles the conversion of relevant columns of the oed files to another currency to do so you will need to provide information on the currency conversion method in the config or after loading

DictBasedCurrencyRates

DictBasedCurrencyRates is a solution where all the rate are provided via files and stored internally as a dictionary.

We support csv file (compressed or not) or a parquet file where they will be read as DataFrame. exemple of currency_conversion_json ("source_type": "parquet" if parquet file is used):

{
    "currency_conversion_type": "DictBasedCurrencyRates",
    "source_type": "csv",
    "file_path": "tests/inputs/roe.csv"
}

The expected format is (roe being a float in parquet format):

cur_from,cur_to,roe
USD,GBP,0.85
USD,EUR,0.95
GBP,EUR,1.12

Rate can also be passed directly in currency_conversion_json ex:

{
    "currency_conversion_type": "DictBasedCurrencyRates",
    "source_type": "dict",
    "currency_rates": [["USD", "GBP", 0.85],
                       ["USD", "EUR", 0.95],
                       ["GBP", "EUR", 1.12]
                      ]
}

When looking for a key pair, DictBasedCurrencyRates check 1st for the key pair (cur1, cur2) then for (cur2, cur1). So if a Currency pairs is only specified one way (ex: GBP=>EUR) then it is automatically assume that roe EUR=>GBP = 1/(roe GPB=>EUR)

if a currency pair is missing ValueError(f"currency pair {(cur_from, cur_to)} is missing") is thrown

FxCurrencyRates

OasisLMF let you use the external package forex-python to perform the conversion. A date may be specified in ISO 8601 format (YYYY-MM-DD) currency_conversion_json:

{
  "currency_conversion_type": "FxCurrencyRates",
  "datetime": "2018-10-10"
}

those config can be added as a json file path of directly into the oed_config dict

config_with_currency_rate = {
    'location': 'SourceLocOEDPiWind.csv', # csv file
    'currency_conversion': {
        "currency_conversion_type": "DictBasedCurrencyRates",
        "source_type": "dict",
        "currency_rates": {
            ('USD', 'GBP'): 0.85,
            ('USD', 'EUR'): 0.952,
            ('GBP', 'EUR'): 1.12}
        },
    'reporting_currency': 'USD',
    }

if reporting_currency is specified in the config, the oed file will be converted on load It can also be set once the OedExposure object has been created

import ods_tools
oed_exposure = ods_tools.oed.OedExposure(**config)
oed_exposure.currency_conversion = ods_tools.oed.forex.create_currency_rates(
    {
        "currency_conversion_type": "DictBasedCurrencyRates",
        "source_type": "dict",
        "currency_rates": {
            ('USD', 'GBP'): 0.85,
            ('USD', 'EUR'): 0.952,
            ('GBP', 'EUR'): 1.12}
        }
)
oed_exposure.reporting_currency = 'EUR' # this line will trigger currency conversion

Format Conversions

The transform command can be used to convert between OED and other formats. To run transformations, and extra set of packages must be installed. This can be done using pip install ods-tools[extra] (or pip install --upgrade ods-tools[extra] if already installed).

Basic csv conversion

A simple csv-to-csv transformation can be run from the command line with

ods_tools transform -f air-oed --input-file source.csv --output-file output.csv

The flag -f instructs the tool on which are the origin and destination formats (currently air and oed).

Complex conversions

More complex transformations run with ods_tools transform requires a configuration file, passed with the option --config-file. Please see the docs here

A configuration file must contain:

  • type of file to transform (location or account)
  • input format
  • output format
  • batch size
  • input file
  • output file

For example:

transformations:
  loc: # type of file to transform, can be "acc" or "loc"
    input_format:
      name: Cede_Location
      version: 10.0.0
    output_format:
      name: OED_Location
      version: 3.0.2
    runner:
      batch_size: 150000 # Number of rows to process in a single batch
    extractor:
      options:
        path: /path/to/input.csv # Path to the input file
        quoting: minimal
    loader:
      options:
        path: /path/to/output.csv # Path to the output file
        quoting: minimal

For a transformation to be run, the folder ./ods_tools/odtf/data/mappings must contain the appropriate yaml configuration file that describes the transformations to perform on the input file to obtain the output (and, potentially, vice versa). Currently, only mapping_loc_Cede-OED.yaml and mapping_acc_Cede-OED.yaml are provided; they describe the transformation between AIR Cede v10.0.0 and OED 3.0.2 for location and account files respectively.

The transformation can be run using:

ods_tools transform --config-file configuration.yaml

Options

The following options can be used to adjust the process:

--nocheck to skip the oed validation at the end of the conversion

-v to adjust the logging level (debug:10, info:20, warning:30, error:40, critical:50)