Checking R list-element names across an Arrow and Parquet round trip
Preserved table metadata does not establish preservation of the names that identify list elements.
Checking R list-element names across an Arrow and Parquet round trip
Inspect the level at which names and attributes give an array meaning. Preserving table-level and column-level metadata does not approve preservation of a named vector inside a list column. Losing the identifiers used by the recipient can change the meaning of otherwise identical values. [1, 2, 3]
Scope of the source report
Arrow issue 54682 presents a round trip retaining R data.frame and ordinary-column attributes but losing names and attributes of list elements. Its author describes losing named-vector identifiers when writing Parquet. This is the reported difference between metadata levels, not a conclusion about every possible R-object attribute. [1]
What to inspect in the input
Extract attributes at table, column and list-element levels, compare their round-trip values and hold files that lost element meaning. Align row identity and list position while distinguishing attribute level and key. If essential names become explicit columns or a sidecar linked by dataset ID, verify that recipients can reconstruct that relationship. [1, 3]
A small acceptance example
Define an illustrative element name as sample_A. Retaining sample_A passes name preservation. An unchanged value array with name state missing loses the basis for distinguishing its sample. Here missing is inspection-table notation for absence, not a literal name. Do not impose this failure if names are irrelevant to the consumer and explicitly outside the preservation contract. [1, 3]
The boundary passed to Truthound
ColumnPairValidator in Truthound Core 3.3.2 can compare extracted before and after attribute values. It skips null pairs, so check name presence separately. Safe extraction of element metadata and R-object reconstruction remain external operations. The validator does not guarantee serialization or restoration of every R attribute. [1, 2, 3]
Application conditions
Link the required original list-element names and restored names by dataset and element position, recording missing names separately.
Alternatives
Normalize required element names into explicit columns and link sidecar metadata by the same dataset identity.
Limitations
This does not guarantee serialization of every arbitrary R object attribute.
Evidence
- Event checked
- Subject checked
- Modified
