Summary
When reading covid-world-vaccination-progress.vortex, a column that should be signed i32 is decoded as unsigned u32, causing large negative values to appear as their unsigned equivalent.
Observed
Expected (oracle): -2146745219 (signed i32)
Got (vortex): 2148222077 (unsigned u32 = -2146745219 + 2^32)
The two values share the same 32-bit bit pattern — this is purely a signed vs unsigned interpretation error.
Detected at CSV line 193132. The affected column is at index 6 (0-based) in the CSV row:
World,OWID_WRL,19182,12148486168,5244055396,4805969879,**-2146745219**,6911415,7438197,152.33,...
Likely cause
Either:
- The column's
DType in the Vortex footer specifies u32 but the Parquet oracle uses i32 for the same column
- Or the Java reader misreads the DType signedness (e.g. PType lookup off-by-one in
PType.fromOrdinal)
Corpus file
~/.cache/raincloud/v1/covid-world-vaccination-progress/vortex/covid-world-vaccination-progress.vortex
Summary
When reading
covid-world-vaccination-progress.vortex, a column that should be signedi32is decoded as unsignedu32, causing large negative values to appear as their unsigned equivalent.Observed
The two values share the same 32-bit bit pattern — this is purely a signed vs unsigned interpretation error.
Detected at CSV line 193132. The affected column is at index 6 (0-based) in the CSV row:
World,OWID_WRL,19182,12148486168,5244055396,4805969879,**-2146745219**,6911415,7438197,152.33,...Likely cause
Either:
DTypein the Vortex footer specifiesu32but the Parquet oracle usesi32for the same columnPType.fromOrdinal)Corpus file