dcp-tools

Manage and load data to Data Commons Platform instances

A Data Commons Platform instance takes your data as CSVs in a fixed variable-per-row shape, plus a config.json that maps each CSV's columns onto the Data Commons schema. If you're defining your own statistical variables, entities, or properties rather than reusing existing ones, you also need MCF (Meta Content Framework) files describing them. dcp-tools builds and validates that bundle in Python (or from the CLI) and uploads it to Cloud Storage to trigger the platform's ingestion workflow.

This package was published as bblocks-datacommons-tools until version 0.1.1, and imported as bblocks.datacommons_tools. Installing the old distribution now pulls in dcp-tools and redirects those imports with a DeprecationWarning, so existing code keeps working. Update imports to dcp_tools when convenient.

Install

pip install dcp-tools

While 1.0.0 is still in alpha, this is a pre-release, so pip won't find it unless you ask for pre-releases explicitly: pip install --pre dcp-tools.

Or from GitHub:

pip install git+https://github.com/ONEcampaign/dcp-tools

Quickstart

This builds a config and MCF for a source, a provenance, one input file, and one statistical variable, then exports the bundle to disk.

from pathlib import Path

import pandas as pd
from dcp_tools import CustomDataManager

manager = CustomDataManager()
manager.add_source(dcid="ONEData", url="https://data.one.org")
manager.add_provenance(
    dcid="ONEClimateFinance",
    url="https://datacommons.one.org/data/climate-finance-files",
    source="ONEData",
)

data = pd.DataFrame({
    "country": ["Kenya", "Kenya", "Vietnam"],
    "year": [2022, 2023, 2023],
    "variable": ["climateFinanceProvidedCommitments"] * 3,
    "value": [12.4, 15.1, 8.7],
})
manager.add_input_file(
    file_name="climate_finance/one_cf_provider_commitments.csv",
    provenance="ONEClimateFinance",
    data=data,
    column_mappings={
        "observationAbout": "country",
        "date": "year",
        "variable": "variable",
        "value": "value",
    },
    observation_properties={"unit": "USDollar"},
)

manager.add_variable_to_mcf(
    dcid="climateFinanceProvidedCommitments",
    name="Climate finance commitments (bilateral)",
    description="Funding committed for climate adaptation and mitigation projects",
    stat_type="dcid:measuredValue",
)

manager.add_mcf_file("*.mcf", provenance="ONEClimateFinance")

out_dir = Path("export/climate_finance")
out_dir.mkdir(parents=True, exist_ok=True)
manager.export_all(out_dir)

export_all writes config.json, the CSV, and both MCF files under out_dir. Since we never called set_import_name, config.json defaults importName to the export directory's name, and column mappings and the provenance name are resolved to full dcids:

{
    "importName": "climate_finance",
    "inputFiles": [
        {
            "filename": "climate_finance/one_cf_provider_commitments.csv",
            "provenance": "dcid:provenance/ONEClimateFinance",
            "columnMappings": {
                "dcid:variableMeasured": "variable",
                "dcid:observationDate": "year",
                "dcid:value": "value",
                "dcid:observationAbout": "country"
            },
            "observationProperties": {"unit": "USDollar"},
            "format": "variablePerRow"
        },
        {
            "pattern": "*.mcf",
            "provenance": "dcid:provenance/ONEClimateFinance"
        }
    ]
}

Loading it

Once you have a bundle on disk, dcp_tools.gcp_utilities uploads it and triggers ingestion:

from dcp_tools.gcp_utilities import (
    get_kg_settings,
    run_ingestion_workflow,
    upload_to_cloud_storage,
)

settings = get_kg_settings(source="env", env_file="customDC.env")
upload_to_cloud_storage(settings=settings, directory="export/climate_finance")
run_ingestion_workflow(settings=settings)

run_ingestion_workflow triggers the DCP (Data Commons Platform) ingestion workflow, which ingests the new data and serves it. There's no separate redeploy step to run. See the loading-data docs for the full settings reference, and the dcp-tools CLI (upload, ingest, pipeline), which wraps this same flow.

Contributing

Contributions are welcome! See CONTRIBUTING for how to get started, report bugs, and submit changes.