Skip to content

Latest commit

 

History

History
508 lines (352 loc) · 20 KB

README.md

File metadata and controls

508 lines (352 loc) · 20 KB
Raw

cfde-deriva

Collaboration point for miscellaneous CFDE-deriva scripts, adapting CFDE concepts to a DERIVA-based hosting platform.

Python package

This repository includes the sources for the cfde_deriva Python package, which is an SDK to assist with various CFDE tasks. It has several useful modules:

These client-side APIs provide CFDE-specific abstractions over the stock deriva-py client SDK which provides generic access to any deriva platform instance, regardless of deployed model.

Configuration data

The preceding Python package includes built-in package data for use with the code. These are further organized by purpose and included as standalone data files of various types.

Submission configuration

The cfde_deriva.configs.submission sub-package includes configuration data related to ingestion of a C2M2 datapackage from BDBag archive format into a SQLite database.

The main datapackage schema product c2m2-datapackage.json is a copy of the C2M2 model definitions, with minor modifications.

Several .tsv files document controlled vocabulary terms which we load as tabular data to support the portal UI. These are represented as enumerations in the C2M2 model definition used by a submitter, but turned back into fields governed by foreign key constraints.

Portal-Prep configuration

The cfde_deriva.configs.portal_prep sub-package includes configuration data related to SQL-based transformations to renormalize the C2M2 content into a different form more appropriate for use in the portal.

The main datapackage schema product cfde-portal-prep.json is derived from the C2M2 model but introduces new supporting tables and renormalizes other columns. Most significantly:

  1. a numeric nid surrogate key is introduced to tables
  2. a core_fact table represents equivalence classes of C2M2 entities with common metadata properties
  3. all foreign keys are replaced by numeric surrogate key references
  4. several transitive-closure tables are computed over existing C2M2 association tables
  5. curated vocabulary and slim-maps from the CFDE Ontology WG are loaded from TSV

Numerous .sql files encode transformations to either compute rows of a derived table or compute additional columns on a table.

Portal configuration

The cfde_deriva.configs.portal sub-package includes configuration data related to the DERIVA-based web portal serving C2M2 inventory information.

The main datapackage schema product cfde-portal.json is derived from the Portal-Prep configuration, effectively describing a subset of the transformation results that will be loaded into the portal's backing database. This configuration also includes numerous DERIVA UI presentation hints to customize how the CFDE portal formats data on screen and provides search and navigation functions.

Registry configuration

The cfde_deriva.configs.registry sub-package includes configuration data related to building a CFDE submission registry, which tracks C2M2 datapackages submitted by participating DCCs.

Command-line tools for Development and Operations

Submission processing

Normally, the cfde-submit CLI tool is used by submitters to introduce new C2M2 datapackages into the submission system. This invokes a workflow system which eventually executes the cfde_deriva.submission.Submission.ingest method to apply the various steps of ingest, provisioning, and configuration implied by the above configuration artifacts.

For CFDE developers and operations staff, a low-level CLI wrapper provides access to directly execute this code from an appropriate development or data preparation workstation, while bypassing the normal hosted workflow system. In order to succeed, the developer or admin user must provide credentials which include sufficient write-access privileges in the CFDE-CC system under test.

The cfde_deriva.submission CLI is primarily a test stub and takes a few settings via environment variables:

  • DERIVA_SERVERNAME=app.example.org (defaults to the dev sandbox API endpoint)
  • CFDE_SKIP_FRICTIONLESS=true will bypass frictionless package validation (default false)
  • CFDE_SKIP_SQL_CHECKS=true will bypass additional relational data checks (default {})
    • general syntax is a JSON object { "check_name": true, ...} to bypass specific checks
    • default {} does not bypass any checks
    • true is a equivalent to an object mapping all check names to true
  • SQLITE_TMPDIR=/path/to/dir will override the default SQLite temp file storage location

Submission emulation

This test-stub exercises cfde-deriva in a partial system test scenario, without relying on the hosted workflow system.

python -m 'cfde_deriva.submission' submit <dcc_id> <archive_url>

Command-line arguments:

  • dcc_id is a CFDE.dcc.id key from the registry
  • archive_url is a BDBag location accessible to the running tool

NOTE: This emulated submission does not write to the datapackage archive system. The input BDBag must be placed in an accessible archive location manually by the developer, prior to using this test interface.

A full system test requires having the new cfde-deriva package deployed to the hosted workflow system (action provider) and using the cfde-submit CLI to start a normal ingest flow.

Submission rebuild

When a submission already exists in the registry, e.g. in a failed state, the developer can restart the idempotent ingest process to attempt to diagnose a software or hosting issue and/or to test new features of the ingest process.

python -m 'cfde_deriva.submission' rebuild <submission_id>

Command-line arguments:

  • submission_id is a CFDE.datapackage.id key from the registry

The developer may wish to intervene with manual changes to the registry and/or ERMrest catalog state of the hosting environemt to clear the way for certain submission build test scenarios.

Submission reconfiguration

When a submission has already been ingested into a data review catalog, a cheaper process can be applied to test revisions to the cfde_deriva.configs.portal UI presentation hints.

python -m 'cfde_deriva.submission' reconfigure <submission_id>

Command-line arguments:

  • submission_id is a CFDE.datapackage.id` key from the registry

This process will only amend annotations and/or access control policies in the already-built ERMrest catalog in order to influence how content is presented through the web UI.

A batch form can reconfigure all existing review catalogs when more test coverage is desired on various datapackage content scenarions:

`python -m 'cfde_deriva.submission' reconfigure-all

It is the operator's responsibility to determine whether this reconfiguration is appropriate, i.e. that existing review catalog content is schema-compatible with the configurations being applied. Older review catalogs may need to be reconfigured by an older, compatible version of cfde-deriva or conversely purged and rebuilt with the submission rebuild sub-command to upgrade to the internal portal schema used by the newer software.

Submission rebuild all in release

A set of constituent submissions can be rebuilt in bulk by a helper routine in the release submodule.

python -m cfde_deriva.release rebuild-submissions <release_id>

command-line arguments:

  • release_id is a CFDE.release.id key from the registry

Submission purge

An administrator can purge ERMrest catalog storage resources associated with a submission, reducing a submission record to basic historical information. (Such a submission can be reconstructed later using the submission rebuild ... command above.)

python -m cfde_deriva.submission purge <submission_id>

Command-line arguments:

  • submission_id is a CFDE.datapackage.id` key from the registry

This purge method can be applied in bulk to many submissions with two special commands.

python -m cfde_deriva.submission purge-auto

python -m cfde_deriva.submission purge-ALL

The purge-auto variant uses heuristics to identify older submissions which are likely irrelevant and purges them. The purge-ALL variant purges every submission, useful in certain administrative scenarios.

Release preparation

When a set of submissions have been ingested, reviewed, and marked "approved" by the appropriate CF Program staff, a new release can be prepared using the cfde_deriva.release CLI wrapper.

This process is relatively expensive as it, in effect, performs ingest over again for each constituent datapackage. However, instead of producing a number of per-submission review catalogs, it produces a merged release catalog, combining all the constituent submission information.

Defining or maintaining the CFDE next-release draft

The release tools facilitate the preparation of a special draft release which is intended to track the likely consituent submissions for the "next CFDE release". When invoked without other arguments, the release draft sub-command idempotently creates or updates this record to reflect the latest approved submissions from each DCC:

python -m cfde_deriva.release draft

A built-in description value (currently future release candidates) is applied to mark the new draft release, and/or used to locate the existing record. Each time the command is invoked, the existing record will be updated to reflect the latest approved submissions.

Environment variables (some shared with the submission module):

  • DERIVA_SERVERNAME=hostname (defaults to the dev sandbox API endpoint)
  • DRAFT_NEED_DCC=true whether DCC approval is required for constituent datapackages (default true)
  • DRAFT_NEED_CFDE=false whether CFDE-CC approval is required for consituent datapackages (default false)

(A draft release is one in the cfde_registry_rel_status:planning state. The implicit next-release draft is both in this state and bearing the special built-in descrption value. In the event that an admin user manually creates multiple draft releases with the same description field value, this tool will locate and maintain the earliest draft, sorting by record creation time in the release registry.)

Defining ad-hoc draft releases

The release tools are oriented around release definitions which are created in the registry before any release data is prepared. Aside from the CFDE next-release draft, the tool can also create other drafts with different description values:

python -m 'cfde_deriva.release' draft new [<description>]

With this invocation, a new CFDE.release record is created in the registry, automatically populated by the most recently submitted and approved release for each CF Program. When a release has already been drafted as such, but you wish to update it to reflect recent submission or approval activities, the previously generated release_id is supplied in place of the new keyword:

python -m 'cfde_deriva.release' draft <release_id> [<description>]

command-line arguments:

  • release_id is a CFDE.release.id key from the registry
  • description is a short, human-readable string which will be visible in the release registry

A release definition can also be created or revised manually via the registry's web UI. This CLI wrapper is a convenience layer to automatically update the stored definitions.

See previous sub-section for environment variables affecting the release draft sub-command.

Release build

A release defined in the registry in a draft/pending state can be built into a catalog using:

python -m 'cfde_deriva.release' build <release_id>

command-line arguments:

  • release_id is a CFDE.release.id key from the registry

The successful build will result in a new catalog existing on the server, without affecting the data currently visible as the de facto public release. Each built catalog exists at its own distinct catalog id in the service.

Normally, a build should be from a clean slate, implicitly provisioning the catalog, ingesting all constitent datapackages, and loading the merged database content into the new catalog. However, on failure the build command can be restarted to attempt idempotent rebuild. Progress markers persistend periodically during the build process are used to efficiently skip ahead to the next task, while underlying idempotent operations are used to allow some work after the last progress marker to be repeated.

Environment variables:

  • DERIVA_SERVERNAME=hostname (defaults to the dev sandbox API endpoint)
  • SQLITE_TMPDIR=/path/to/dir will override the default SQLite temp file storage location
  • REPROVISION_MODEL=false whether to attempt to apply model changes to a build that is in progress

Normally, restart should use the exact same cfde-deriva codebase and so all model provisioning occurs once and only once. However, for development and test purposes, incremental reprovisioning can be enabled to attempt to more rapidly test some small changes, such as the addition of another computed table or column. However, it is the developer's responsibility to decide when this short-cut is valid, and to start over with a clean build when circumstances require it.

Release reconfiguration

A release that is already built can be updated with display hints or other non-data configuration changes, without rebuilding the tabular data content.

python -m 'cfde_deriva.release' reconfigure <release_id>

command-line arguments:

  • release_id is a CFDE.release.id key from the registry

Release publishing

After a release has been built and inspected by QA users, a quick task can change which catalog is addressed by the CFDE standard release catalog ID of 1. This uses a DERIVA feature known as a catalog alias wherein the special catalog id 1 is an alias bound to a specific release catalog, and the alias can be moved to target newer releases over time.

python -m 'cfde_deriva.release' publish <release_id>

command-line arguments:

  • release_id is a CFDE.release.id key from the registry

After moving the catalog alias 1 to point to the specified release's catalog, this tool also attempts to update release status metadata in the registry to document the change:

  1. Change the new release to "public-release" state
  2. Replace the built-in next-release description with a release date description
  3. Change the previous release to "obsoleted" state

If used in unexpected ways, the tool may move the alias but skip metadata updates, requiring the CFDE admin users to manually correct release status via the web UI.

Release prune favorites

The user-specific favorite vocabulary terms can be pruned to remove orphaned terms by using a given release as the reference source for known terms.

python -m 'cfde_deriva.release' prune-favorites <release_id>

  • release_id is a CFDE.release.id key from the registry

Release refresh resource markdown

The vocabulary term resource markdown content in a release can be refreshed to match the latest in the registry.

python -m cfde_deriva.release refresh-resources <release_id>

Command-line arguments:

  • release_id is a CFDE.release.id` key from the registry

The release must already be in a content-ready or released state to allow refresh.

Release purge

An administrator can purge ERMrest catalog storage resources associated with a release, reducing a release record to basic historical information. (Such a release can be reconstructed later using the release build ... command above.)

python -m cfde_deriva.release purge <release_id>

Command-line arguments:

  • release_id is a CFDE.release.id` key from the registry

This purge method can be applied in bulk to many releases with two special commands.

python -m cfde_deriva.release purge-auto

python -m cfde_deriva.release purge-ALL

The purge-auto variant uses heuristics to identify older releases which are likely irrelevant and purges them. The purge-ALL variant purges every release, useful in certain administrative scenarios.

Registry maintenance

The registry is a long-lived component of the CFDE portal infrastructure. However, similar tools are available to perform maintenance tasks including the provisioning of a brand-new system.

Command-line arguments shared by nearly all registry sub-commands:

  • catalog_id is an ERMrest catalog ID string (default registry`)

If provided, an alternate catalog ID allows for testing with a parallel instance of the registry, independent of the normal CFDE registry maintained with the default ID registry.

Provision registry

A new registry can be created from scratch as a new ERMrest catalog which will be initialized with the registry model and initial, default content.

python -m cfde_deriva.registry provision [<catalog_id>]

It is an error to provision a registry with a catalog ID that already exists in the ERMrest service host.

Reprovision registry

An existing registry can be upgraded, i.e. schema-migrated.

python -m cfde_deriva.registry reprovision [<catalog_id>]

unlike the previous command, this requires the registry catalog to already exist, so it performs incremental upgrades.

Reconfigure registry

An existing registry can also be given configuration changes, without model or data updates.

python -m cfde_deriva.registry reconfigure [<catalog_id>]

Delete registry

An existing registry can be deleted, e.g. to remove a parallel test registry.

python -m cfde_deriva.registry delete [<catalog_id>]

WARNING: care must be taken as this command, if misused, can also remove other ERMrest catalogs whether or not they were actually produced by the registry module.

Fixup FQDN

If a CFDE service host is moved from one fully-qualified domain name to another, e.g. due to changes in a load balancer config or DNS records in the hosting environment, the registry UX will be internally-inconsistent due to URLs stored in the registry records. A fixup command rewrites these URLs to match the current service host FQDN.

python -m cfde_deriva.registry fixup-fqdn [<catalog_id>]

Upload resource markdown

Curated, markdown-formatted resource information can be uploaded to augment vocabulary terms known by the registry.

python -m cfde_deriva.registry upload-resources vocabname.json...

One or more JSON files are supplied and must be named by a vocabulary table name known by the registry plus the suffix .json. For example:

  • anatomy.json
  • assay_type.json
  • ethnicity.json

This list is not exhaustive, so see other C2M2 documentation for more vocabulary names.

The structure of each input JSON file is an array of simple records:

[
  {"id": "CURI", "resource_markdown": "Resource description using **markdown**."},
  ...
]

Each record identifies one existing term by its CURI id and supplies the extra resource_markdown payload. Markdown often contains newlines and these will be escaped in the JSON payload as \n, a backslash character followed by the character n.

As soon as uploads are completed, the results can be reviewed by browsing the C2M2 vocabulary tables in the submission system itself.