OCDS Commands

Optional arguments for all commands are:

  • --encoding ENCODING the file encoding
  • --ascii print escape sequences instead of UTF-8 characters
  • --pretty pretty print output
  • --root-path ROOT_PATH the path to the items to process within each input

The inputs can be concatenated JSON or JSON arrays.

Note

An error is raised if the JSON is malformed or if the --encoding is incorrect.

Handling edge cases

Large packages

If you are working with individual packages that are too large to hold in memory, use the echo command to reduce their size.

Embedded data

If you are working with files that embed OCDS data, use the --root-path ROOT_PATH option to indicate the “root” path to the items to process within each input. For example, if release packages are in an array under a results key, like so:

{
  "results": [
    {
      "uri": "placeholder:",
      "publisher": {"name": ""},
      "publishedDate": "9999-01-01T00:00:00Z",
      "version": "1.1",
      "releases": []
    }
  ]
}

You can run ocdskit <command> --root-path results to process the release packages. The root path, in this case, is simply the results key. OCDS Kit will read the entire results array into memory, and process each array entry.

If the results array is very large, you should run ocdskit <command> --root-path results.item instead. The root path, in this case, is the results key joined to the item literal by a period (the item literal indicates that the items to process are in an array). OCDS Kit will read each array entry into memory, instead of the entire results array.

For this next example, you can run ocdskit <command> --root-path results.item.ocdsReleasePackage:

{
  "results": [
    {
      "ocdsReleasePackage": {
        "uri": "placeholder:",
        "publisher": {"name": ""},
        "publishedDate": "9999-01-01T00:00:00Z",
        "version": "1.1",
        "releases": []
      }
    }
  ]
}

The root path, in this case, is the results key joined to the item literal, joined to the ocdsReleasePackage key.

detect-format

Reads OCDS files, and reports whether each is:

  • a release package
  • a record package
  • a release
  • a record
  • a compiled release
  • a versioned release
  • a JSON array of one of the above
  • concatenated JSON of one of the above

Mandatory positional arguments:

  • file OCDS files
ocdskit detect-format tests/fixtures/realdata/release-package-1.json tests/fixtures/realdata/record-package-1.json

compile

Reads release packages and individual releases from standard input, merges the releases by OCID, and prints the compiled releases.

Optional arguments:

  • --schema SCHEMA the URL or path of the release schema to use
  • --package wrap the compiled releases in a record package
  • --linked-releases if --package is set, use linked releases instead of full releases, if the input is a release package
  • --versioned if --package is set, include versioned releases in the record package; otherwise, print versioned releases instead of compiled releases
  • --uri URI if --package is set, set the record package’s uri to this value
  • --published-date PUBLISHED_DATE if --package is set, set the record package’s publishedDate to this value
  • --version VERSION if --package is set, set the record package’s version to this value
  • --publisher-name PUBLISHER_NAME if --package is set, set the record package’s publisher’s name to this value
  • --publisher-uri PUBLISHER_URI if --package is set, set the record package’s publisher’s uri to this value
  • --publisher-scheme PUBLISHER_SCHEME if --package is set, set the record package’s publisher’s scheme to this value
  • --publisher-uid PUBLISHER_UID if --package is set, set the record package’s publisher’s uid to this value
  • --fake if --package is set, set the record package’s required metadata to dummy values

If --package is set, and if the --publisher-* options aren’t used, the output package will have the same publisher as the last input package.

cat tests/fixtures/realdata/release-package-1.json | ocdskit compile > out.json

For the Python API, see ocdskit.combine.merge().

Note

An error is raised if a release is missing an ocid field, or if the values of the release packages’ version fields are inconsistent.

upgrade

Upgrades packages, records and releases from an old version of OCDS to a new version. Any data not in the old version is passed through. Note: Versioned releases within a record package are not upgraded.

OCDS 1.0 describes an organization’s name, identifier, address and contactPoint as relevant to identifying it. OCDS 1.1 moves organization data into a parties array. To upgrade from OCDS 1.0 to 1.1, we create an id for each organization, based on those identifying fields. This can result in duplicates in the parties array, if the same organization has different or missing values for identifying fields in different contexts. This can also lead to data loss if the same organization has different values for non-identifying fields in different contexts; the command prints warnings in such cases.

Note: OCDS 1.0 uses the whole-list merge strategy on the suppliers array to prepare the compiled release and versioned release, whereas OCDS 1.1 uses the identifier merge strategy. This means that you should merge first and then upgrade.

Mandatory positional arguments:

  • versions the colon-separated old and new versions
cat tests/fixtures/realdata/release-package-1.json | ocdskit upgrade 1.0:1.1 > out.json

For the Python API, see Upgrade.

If a release package is too large, you can upgrade its individual releases using --root-path releases.item.

Note

An error is raised if upgrading between the specified versions is not implemented.

package-records

Reads records from standard input, and prints one record package.

Optional positional arguments:

  • extension add this extension to the package

Optional arguments:

  • --uri URL set the record package’s uri to this value
  • --published-date PUBLISHED_DATE set the record package’s publishedDate to this value
  • --version VERSION set the record package’s version to this value
  • --publisher-name PUBLISHER_NAME set the record package’s publisher’s name to this value
  • --publisher-uri PUBLISHER_URI set the record package’s publisher’s uri to this value
  • --publisher-scheme PUBLISHER_SCHEME set the record package’s publisher’s scheme to this value
  • --publisher-uid PUBLISHER_UID set the record package’s publisher’s uid to this value
  • --fake set the record package’s required metadata to dummy values
cat tests/fixtures/record_*.json | ocdskit package-records > out.json

To convert record packages to a record package, you can use the --root-path option:

cat tests/fixtures/realdata/record-package* | ocdskit package-records --root-path records.item

If --uri and --published-date are not set, the output package will be invalid. Use --fake to set placeholder values.

For the Python API, see ocdskit.combine.package_records().

package-releases

Reads releases from standard input, and prints one release package.

Optional positional arguments:

  • extension add this extension to the package

Optional arguments:

  • --uri URL set the release package’s uri to this value
  • --published-date PUBLISHED_DATE set the release package’s publishedDate to this value
  • --version VERSION set the release package’s version to this value
  • --publisher-name PUBLISHER_NAME set the release package’s publisher’s name to this value
  • --publisher-uri PUBLISHER_URI set the release package’s publisher’s uri to this value
  • --publisher-scheme PUBLISHER_SCHEME set the release package’s publisher’s scheme to this value
  • --publisher-uid PUBLISHER_UID set the release package’s publisher’s uid to this value
  • --fake set the release package’s required metadata to dummy values
cat tests/fixtures/release_*.json | ocdskit package-releases > out.json

To convert record packages to a release package, you can use the --root-path option:

cat tests/fixtures/realdata/record-package* | ocdskit package-releases --root-path records.item.releases.item

If --uri and --published-date are not set, the output package will be invalid. Use --fake to set placeholder values.

For the Python API, see ocdskit.combine.package_releases().

combine-record-packages

Reads record packages from standard input, collects packages and records, and prints one record package.

If the --publisher-* options aren’t used, the output package will have the same publisher as the last input package.

Optional arguments:

  • --uri URL set the record package’s uri to this value
  • --published-date PUBLISHED_DATE set the record package’s publishedDate to this value
  • --version VERSION set the record package’s version to this value
  • --publisher-name PUBLISHER_NAME set the record package’s publisher’s name to this value
  • --publisher-uri PUBLISHER_URI set the record package’s publisher’s uri to this value
  • --publisher-scheme PUBLISHER_SCHEME set the record package’s publisher’s scheme to this value
  • --publisher-uid PUBLISHER_UID set the record package’s publisher’s uid to this value
  • --fake set the record package’s required metadata to dummy values
cat tests/fixtures/record-package_*.json | ocdskit combine-record-packages > out.json

If you need to create a single package that is too large to hold in your system’s memory, please comment on this issue.

For the Python API, see ocdskit.combine.combine_record_packages().

Note

A warning is issued if a package’s "records" field isn’t set.

combine-release-packages

Reads release packages from standard input, collects releases, and prints one release package.

If the --publisher-* options aren’t used, the output package will have the same publisher as the last input package.

Optional arguments:

  • --uri URL set the release package’s uri to this value
  • --published-date PUBLISHED_DATE set the release package’s publishedDate to this value
  • --version VERSION set the release package’s version to this value
  • --publisher-name PUBLISHER_NAME set the release package’s publisher’s name to this value
  • --publisher-uri PUBLISHER_URI set the release package’s publisher’s uri to this value
  • --publisher-scheme PUBLISHER_SCHEME set the release package’s publisher’s scheme to this value
  • --publisher-uid PUBLISHER_UID set the release package’s publisher’s uid to this value
  • --fake set the release package’s required metadata to dummy values
cat tests/fixtures/release-package_*.json | ocdskit combine-release-packages > out.json

If you need to create a single package that is too large to hold in your system’s memory, please comment on this issue.

For the Python API, see ocdskit.combine.combine_release_packages().

Note

A warning is issued if a package’s "releases" field isn’t set.

split-record-packages

Reads record packages from standard input, and prints smaller record packages for each.

Mandatory positional arguments:

  • size the number of records per package
cat tests/fixtures/realdata/record-package-1-2.json | ocdskit split-record-packages 2 | split -l 1 -a 4

The split command will write files named xaaaa, xaaab, xaaac, etc. Don’t combine the OCDS Kit --pretty option with the split command.

split-release-packages

Reads release packages from standard input, and prints smaller release packages for each.

Mandatory positional arguments:

  • size the number of releases per package
cat tests/fixtures/realdata/release-package-1-2.json | ocdskit split-release-packages 2 | split -l 1 -a 4

The split command will write files named xaaaa, xaaab, xaaac, etc. Don’t combine the OCDS Kit --pretty option with the split command.

tabulate

Reads packages, records or releases from standard input and stores releases in a relational database.

Mandatory positional arguments:

Optional arguments:

  • --drop drop all tables before loading
  • --schema SCHEMA the release-schema.json to use
cat release_package.json | ocdskit tabulate sqlite:///data.db

For the format of database_url, see the SQLAlchemy documentation.

Database structure

The database structure follows the specified schema. By default, the latest OCDS release schema is used.

The primary table is the releases table. Secondary tables are created for each array in the schema: for example, there is an awards table and an awards_suppliers table.

Naming conventions

Table names are based on the JSON paths to the arrays, separated by underscores. For example, the data in the /contracts/*/implementation/documents array is stored in the contracts_implementation_documents table.

Column names are based on the JSON paths to the fields, relative to their containing arrays, separated by underscores. For example, the data in the /parties/*/address/region field is stored in the address_region column of the parties table.

Foreign keys

Foreign keys are used to JOIN rows that relate to the same object across tables.

Every table has an ocid column to identify the contracting process, and a release_id column to identify the release.

Secondary tables have additional foreign keys to identify a specific object in a given array. The column names follow the pattern {singular}_id, where singular is all but the last character of the table name. For example:

  • The awards table has an award_id column to identify the award object by its /awards/*/id value.
  • The awards_suppliers table has an award_id column to identify the award object by its /awards/*/id value, and a supplier_id column to identify the supplier by its /awards/*/suppliers/*/id value.

Additional fields

Fields in the JSON data that aren’t described by the provided schema are treated as follows:

  • If the field is an array, it is ignored and a warning is reported, for example: table tender_participationFees does not exist.
  • Otherwise, it is stored in a JSON object in an extras column. For example, an /awards/*/exchangeRate value is stored in the extras column of the awards as a JSON object like {"exchangeRate": 1.23}.

To limit the number of fields that are stored in the extras column, extend the release schema with all relevant extensions, and then use the --schema option.

Alternative approaches

Kingfisher Process stores OCDS releases as JSON blobs in a single column.

Flatten Tool flattens JSON data into CSV and Excel files and supports additional fields, additional arrays and many other ways to customize the output.

validate

Reads JSON data from standard input, validates it against the schema, and prints errors.

Optional arguments:

  • --schema SCHEMA the URL or path of the schema to validate against
  • --check-urls check the HTTP status code if “format”: “uri”
  • --timeout TIMEOUT timeout (seconds) to GET a URL
  • --verbose print items without validation errors
cat tests/fixtures/* | ocdskit validate

echo

Repeats the input, applying --encoding, --ascii, --pretty and --root-path, and using the UTF-8 encoding.

You can use this command to reformat data:

  • Use UTF-8 encoding:

    cat iso-8859-1.json | ocdskit --encoding iso-8859-1 echo > utf-8.json

  • Use ASCII characters only:

    cat unicode.json | ocdskit --ascii echo > ascii.json

  • Use UTF-8 characters where possible:

    cat ascii.json | ocdskit echo > unicode.json

  • Pretty print:

    cat compact.json | ocdskit --pretty echo > pretty.json

  • Make compact:

    cat pretty.json | ocdskit echo > compact.json

You can also use this command to extract releases from release packages, and records from record packages. This is especially useful if a single package is too large to hold in memory.

  • Split a large record package into smaller packages of 100 records each:

    cat large-record-package.json | ocdskit echo --root-path records.item | ocdskit package-records --size 100

  • Split a large release package into smaller packages of 1,000 releases each:

    cat large-release-package.json | ocdskit echo --root-path releases.item | ocdskit package-releases --size 1000

Note that the package metadata from the large package won’t be retained in the smaller packages; you can use the optional arguments of the package-records and package-releases commands to set the package metadata.

If the single package is small enough to hold in memory, you can use the split-record-packages and split-release-packages commands instead, which retain the package metadata.

convert-to-oc4ids

Reads individual releases or release packages from standard input, and prints a single project conforming to the Open Contracting for Infrastructure Data Standards (OC4IDS). It assumes all inputs belong to the same project.

The logic for the mappings between OCDS and OC4IDS fields is documented here.

Optional arguments:

  • --project-id PROJECT_ID set the project’s id to this value
  • --all-transforms run all optional transforms
  • --transforms OPTIONS comma-separated list of optional transforms to run
  • --package wrap the project in a project package
  • --uri URI if --package is set, set the project package’s uri to this value
  • --published-date PUBLISHED_DATE if --package is set, set the project package’s publishedDate to this value
  • --version VERSION if --package is set, set the project package’s version to this value
  • --publisher-name PUBLISHER_NAME if --package is set, set the project package’s publisher’s name to this value
  • --publisher-uri PUBLISHER_URI if --package is set, set the project package’s publisher’s uri to this value
  • --publisher-scheme PUBLISHER_SCHEME if --package is set, set the project package’s publisher’s scheme to this value
  • --publisher-uid PUBLISHER_UID if --package is set, set the project package’s publisher’s uid to this value
  • --fake if --package is set, set the project package’s required metadata to dummy values
cat releases.json | ocdskit convert-to-oc4ids > out.json

Transforms

The transforms that are run are described here.

  • additional_classifications, description, sector, title: populate top-level fields with their equivalents from planning.project
  • administrative_entity, public_authority_role, procuring_entity, suppliers: populate the parties field according to the party role
  • budget: populates budget.amount with its equivalent
  • budget_approval, environmental_impact, land_and_settlement_impact and project_scope: populate the documents field from planning.documents according to the documentType
  • contracting_process_setup: Sets up the contractingProcesses array of objects with id, summary, releases and embeddedReleases. Some of the other transforms depend on this, so it is run first
  • contract_period: populates the summary.contractPeriod field with appropriate values from awards or tender
  • contract_price: populates the summary.contractValue field with the sum of all awards.value fields where the currency is the same
  • cost_estimate: populates the summary.tender.costEstimate field with the appropriate tender.value
  • contract_process_description: populates the summary.description field from appropriate values in contracts, awards or tender
  • contract_status: populates the summary.status field using the contractingProcessStatus codelist.
  • contract_title: populates summary.title from the title field in awards, contracts or tender
  • final_audit: populate the documents field from contracts.implementation.documents according to the documentType
  • funding_sources: updates parties with organizations having funder in their roles or from planning.budgetBreakdown.sourceParty
  • location: populates the locations field with an array of location objects from planning.projects.locations
  • procurement_process: populates the .summary.tender.procurementMethod and .summary.tender.procurementMethodDetails fields with their equivalents from tender
  • purpose: populates the purpose field from planning.rationale

Optional transforms

Some transforms are not run automatically, but only if set. The following transforms are included if they are listed in using the --transforms argument (as part of a comma-separated list) or if --all-transforms is passed.

  • buyer_role: updates the parties field with parties that have buyer in their roles
  • description_tender: populate the description field from tender.description if no other is available
  • location_from_items: populate the locations field from deliveryLocation or deliveryAddress in tender.items if no other is available
  • project_scope_summary: updates summary.tender with items and milestones from tender
  • purpose_needs_assessment: populate the documents field from planning.documents according to the documentType needsAssessment
  • title_from_tender: populate the title field from tender.title if no other is available

Transformation Notes

Most transforms follow the logic in the mapping documentation. However, there is some room for interpretation in some of the mappings, so here are some notes about these interpretations.

Differing text across multiple contracting process

planning/project/title, project/planning/description (planning and budget extension):

If there are any contradictions i.e one contract says the title is different from another a warning is raised and the field is ignored in that case. If all contracting processes agree (when the fields exists in them) then the value is still used.

tender/title, tender/description, /planning/rationale:

If there a multiple contradicting process then we concatenate the strings and put the ocid in angle brackets like:

<someocid> a tender description <anotherocid> another description

If there is only one contracting processes then the ocid part is omitted.

Parties ID across multiple contracting processes

When parties/id from different contracting processes are conflicting and also if there are parties in multiple contracting processes that are the same, we need to identify which are in fact the same party.

The logic that the transforms do to work out matching parties:

  • If all parties/id are unique across contracting processes then do nothing and add all parties to the project.
  • If there are conflicting parties/id then look at the identifier field and if there are scheme and id make an id of somescheme-someid and use that in order to match parties across processes. If there are different roles then add them to the same party. Use the other fields from the first party found with this id.
  • If there is no identifier then make up a new auto increment number and use that as the id. This means the original IDs get replaced and are lost in the mapping
  • If there is no identifier and all fields apart from roles and id are the same across parties then treat that as a single party and add the roles together and use a single generated id.

Document ID across multiple contracting processes

If there are are only unique project/documents/id keep the ids the same. Otherwise create a new auto-increment for all docs. This means the original ``documents/id`` are lost

Project Sector

Sectors are gathered from planning/project/sector and it gets all unique scheme and id of the form <scheme>-<id> and adds them to the sector array. This could mean that the sectors generated are not in the Project Sector Codelist.

Project Scope Summary

If --all-transforms is set or if project_scope_summary is included in --transforms it copies over all tender/items and tender/milestones to contractingProcess/tender. This is to give the output enough information in order to infer project scope.