LoginSign Up
← Back to all sparks
E

eatGADS

INFRA · APIS
Velocity0.0

Data Management of Large Hierarchical Data

github.com/beckerbenj/eatgads ↗

Labelled survey data, one edge case at a time — and the edge cases are all about missing values.

survey-dataspssvalue-labelsmissing-datar-packageeducational-assessment
◆Current state
eatGADS manages labelled survey and assessment data in R, the SPSS-descended world of value labels, missing tags, and metadata that has to survive every transformation. Releases are roughly annual and dense. The current one, 1.2.0, is about doing things in bulk: changing value labels, missing tags, and recodings across many variables at once rather than one at a time.
◆Where it's heading
The feature line and the bug line point at the same thing from opposite directions. Features keep widening the aperture — multiple variables, multiple ID variables, comparisons within a single object — while fixes keep landing on the collision between value labels and missing codes, where a value can be labelled NA, duplicated, or tagged and transformed away. extractData() and extractData2() alone absorbed eight separate correctness fixes across the last two releases. The package is hardening the one place labelled data is most likely to lose information.
◆Prediction
Expect the bulk-operation pattern to spread to the remaining single-variable functions, and continued fixes wherever value labels and missing tags interact; the extraction path is clearly still the weak point.

◆Recent moves

  1. 1y ago

    Value labels, missings, and recodes go multi-variable

    The bulk-operations release: changeValLabels(), changeMissings(), and recodeGADS() all accept multiple variables, and equalData()/equalMeta() split equalGADS() into its data and metadata halves while gaining support for composite ID variables. The fix list is again dominated by extractData() and duplicate or NA-labelled values.

    View source ↗
  2. 1y ago

    Tibble import and within-object difference inspection

    import_tibble() adds a tibble entry point alongside the SPSS import path, and the inspectDifferences() pair can now compare variables inside a single GADSdat rather than only across two. Ten bug fixes underneath, most of them in the same value-label-versus-missing-tag territory the next release also works on.

    View source ↗
  3. 3y ago

    Variable clone, create, insert, and auto-recode round out the toolkit

    The 1.0.0 release rounds out the variable-manipulation toolkit — cloneVariable(), createVariable(), insertVariable(), autoRecode() — so a GADSdat can be restructured in place rather than only imported and checked. extractData2() arrives here with ordered-factor support, and encoding and date handling in import_spss() get real attention.

    View source ↗