LoginSign Up
← Back to all sparks
X

xportr

ANALYTICS
Velocity0.0

Tools to build CDISC compliant data sets and check for CDISC compliance.

github.com/atorus-research/xportr ↗

The CDISC transport writer collapsed six pipeline calls into one, then spent two years hardening it

cdiscclinical-submissionssas-transportpharmaversemetadatabreaking-changes
◆Current state
xportr applies CDISC metadata — variable types, lengths, labels, formats, ordering — to R data frames and writes the SAS transport files that go into regulatory submissions. Since v0.4.0 the package has had a single entry point, xportr_process(), that runs the whole chain and writes, and metadata arrives as a plain specification rather than a metacore object. The v0.5.0 release in January 2026 finished the cleanup by deleting every deprecated argument left over from that redesign.
◆Where it's heading
The work has moved from building the pipeline to defending it against the ways submission data actually arrives: grouped data frames now raise a warning, date and time variables get class checks, illegal characters are resolved rather than erroring, and xportr_write() warns before a file crosses 5GB instead of producing an unusable artifact. Contributor volume is high and spread across sponsors — Atorus, Roche, GSK and others show up in the PR lists — which is what keeps a validated-context package moving without a single owner.
◆Prediction
With deprecations cleared in 0.5.0, the next cycle likely targets more input-shape validation of the kind 0.5.0 started — the grouped-data and datetime-class checks read as the first two of a series. Nothing in these entries points to a new output format.

◆Recent moves

  1. 7mo ago

    Deprecated arguments deleted; 5GB and datetime guards added

    Removes every deprecated argument across the xportr functions, closing out the API migration that v0.4.0 started — code still passing the old arguments breaks here. Alongside it: a warning when an export would exceed 5GB, class checks on date/time variables, a warning on grouped input, an illegal-character fix, and repairs to the verbose flags including a missing xportr.order_verbose option.

    View source ↗
  2. 1y ago

    Test fixes for the metacore 0.2.0 update

    Repairs tests that broke when metacore released v0.2.0, plus a version bump. No change to what xportr produces.

    View source ↗
  3. 1y ago

    hms accepted as a numeric type; domain logging filled in

    Adds hms to xportr.numeric_types so time variables survive the type pass, corrects a misleading message in xportr_order(), and fills in missing domain logging. Small, but the hms handling removes a manual workaround for any dataset carrying time-of-day.

    View source ↗
  4. 1y ago

    File-size limit moves into xportr_write, deprecating xportr_split

    Moves file splitting inside xportr_write() as a maximum-size-in-GB argument and deprecates the standalone xportr_split(). Consistent with the 0.4.0 direction of folding steps into the main path rather than exposing them as separate functions callers have to sequence themselves.

    View source ↗
  5. 2y ago

    xportr_process() runs the whole pipeline in one call

    ⚡ SPARK

    The release that redefines how xportr is used: xportr_process() runs the full metadata chain and writes in one call, xportr_options() manages configuration centrally, xportr_write() takes metadata directly while the metacore argument is hard-deprecated, and lengths derive from actual data. It is the origin of every deprecation that v0.5.0 later deletes.

    View source ↗
  6. 2y ago

    Unused dependencies removed

    Drops unused package dependencies. Housekeeping ahead of the 0.4.0 redesign.

    View source ↗