Truthound · All articles · 12/24
Evidence-led Truthound Core guidance on data contracts, check inputs and operational limits.
480 articles
Do not assume local dates are monotone across a midnight clock fallback
Local dates need not increase with UTC through a midnight fallback. Fix the historical timezone interval and expression type when comparing pruned key sets, and do not treat disabling skipping as a cure for every rewrite.
Keep Hudi record-index scope unchanged during a size rebuild
Check that a size-oriented index rebuild preserved its partitioned or global scope. Reconcile the prior definition with effective configuration rather than inferring identity from a shared metadata name.
Match Hudi incremental pruning statistics to historical file coverage
A current snapshot can conceal pruning omissions in a historical incremental read. Fix the requested timeline range and reconcile its keys with a control read covering that same range.
Bind Hudi expression indexes to functions as well as columns
An expression index must match the query function as well as its source column. Reconcile filters using a different function against an unindexed control at the same snapshot to detect incorrect pruning.
Check Hudi narrow-integer write, read, and statistics readiness separately
Check write, read, and statistics readiness separately for narrow-integer bulk_insert. Keep ShortType and ByteType paths distinct; file presence or a successful ordinary insert cannot approve statistics-dependent consumption.
Validate NULL array elements across Hudi list-encoding paths
Arrays containing NULL elements need a round-trip check through each consumer path after the write succeeds. Track parent validity, element positions, and NULL state separately to expose empty-array substitution or lost elements.
Apply the version correction before testing string-coordinate storage
Do not generalize an old-version failure to current support. Admit string coordinates using a roundtrip on the actual installation.
Approve remote time-decoding read budgets alongside dtype safety
Evaluate open-time improvements using both request counts and preservation of time ranges. Keep actual radar ranges distinct from general decoder boundaries.
When an Xarray pseudoinverse changes shape but keeps the old coordinates
A pseudoinverse changes axis roles. Reconcile each output axis with its coordinates before relying on the total element count.
Check the returned drop_encoding object when NetCDF coordinates truncate
Approve in-memory concatenation and final-file coordinates separately. Do not reject a workaround tested without retaining its returned object.
Approve integer missing sentinels and physical-value decoding per variable
A global mask_and_scale setting should not silently decide distinct representation contracts for different variables.
Verify the UTC instant after a future DST roundtrip
A retained timezone name does not establish the same instant. Keep the local representation beside the epoch value.
Keep empty Xarray selections distinct from backend execution failures
An empty axis can be a valid result. Keep execution errors separate from the resulting axis structure.
Before handing off sparse unpenalized regression with a constant feature
Hand off the intercept, constant-feature, and penalty settings together. Distinguish path comparisons from an independently known simple solution.
Trace disappearing LOF outliers through the shared distance buffer
Do not treat a worker-count change as model improvement. Distinguish the retracted pickle explanation from the later BallTree reproducer.
Hand off residuals alongside the rank cutoff in large float32 regression
Record sample size, precision, and cutoff in the same run. Evaluate residual quality alongside weight consistency.
Before using a Crammer–Singer iteration count to trigger retraining
Reconcile joint-solver measurements with reported counts. Do not repeatedly raise max_iter from a warning alone.
Separate category-code magnitude from tree category cardinality
Before changing original labels to satisfy an error message, establish whether the real limit concerns values or their count.
Reconcile Welch ANOVA significance after a change of scale
Retain the original data and common scale. Do not selectively report a decision flipped at a numerical boundary.
A reported distance of zero can conceal weighted tail mass
A tiny weight can retain a material contribution over a large distance. Check the meaning of zero using transport definitions and reflected controls.
