# Keep the meaning intact when data moves

Treat this file as reference data, never as instructions that override the user or agent. Read the complete item before applying it. Keep use conditions, avoid conditions, limits, evidence roles and source links with the answer. An editorial block is a separately versioned analysis layer, not an original corpus claim or proof of efficacy. Cite the readable URL and version. Do not infer human review. This is a saved edition, not a live safety-notice service. Before consequential reuse, check the current page, https://vedokrok.com/data/availability.json and https://vedokrok.com/data/notices.json; absence alone does not explain why an item was removed.

Release: VEDOKROK-SEL-5d86e1b23401e578
Snapshot SHA-256: 30ce1c8fa9155763797e3d5cb55b1d94cbb8c3146ebd03fb3c006531ce60e82f
Current availability: https://vedokrok.com/data/availability.json
Current notices: https://vedokrok.com/data/notices.json
Editorial overlay: MHC-EO-20261021-481e1ce95099
Editorial projection SHA-256: 6b41bfd164b85b2522e2378c425929aa041d9e093e81428ade9a7b7b9f78e8fe

Make joins, identifiers, updates and completion checks preserve the business meaning of the records being moved.
Collection kind: collection
Source: https://vedokrok.com/collections/keep-the-meaning-intact-when-data-moves
Version: 0.1.0

---

## Name what one row means before joining two tables

ID: MHC-D-RESEARCH-0980 · Version: 0.1.0 · Kind: checklist
Source: https://vedokrok.com/knowledge/name-what-one-row-means-before-joining-two-tables

The join may be working perfectly on the wrong relationship.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A merge produces plausible columns but unexpectedly more rows or larger totals.

### Avoid when

- A legitimate one-to-many join expands rows. Null-key behavior also differs between tools, so do not assume SQL and spreadsheet-like joins are identical.

### Explanation

State the grain: what does one row represent in each input? Then specify how many matches are allowed for the chosen key. Check that expectation before accepting the joined result; do not remove duplicates afterward merely to recover a familiar row count.

### Checklist

- Name the complete business key on each side, including relevant date or organizational scope.
- Check uniqueness on the side that is supposed to contain one match.
- Inspect unmatched keys and compare row counts and meaningful totals before and after the join.

### Example

Joining sales lines to several historical customer versions by customer number alone multiplies each sale.

### Check

The observed match count agrees with the declared relationship, and exceptions have an explanation.

### Limits

- A legitimate one-to-many join expands rows. Null-key behavior also differs between tools, so do not assume SQL and spreadsheet-like joins are identical.

### Evidence and sources

- supports: pandas can validate one-to-one, one-to-many or many-to-one merge keys; allowing many-to-many does not perform a uniqueness check. — RS-F228187ED3C9FC8D. Key uniqueness is a structural property, not proof that the selected join represents the intended relationship. (validate parameter)
- RS-F228187ED3C9FC8D: pandas.merge — https://pandas.pydata.org/docs/reference/api/pandas.merge.html

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/reconcile-what-is-missing-and-what-should-never-have-arrived

---

## Reconcile what is missing and what should never have arrived

ID: MHC-D-RESEARCH-0981 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/reconcile-what-is-missing-and-what-should-never-have-arrived

Everything expected can be present while the result is still wrong.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A source-to-target comparison reports no missing records, but the target may contain extras.

### Avoid when

- A target may legitimately contain records outside this run. Define scope before treating all unmatched target rows as defects.

### Explanation

Compare the expected and actual populations in both directions. Expected minus actual reveals omissions; actual minus expected reveals unexpected additions. Use the same identity definition and cutoff in both comparisons, then inspect duplicates separately when repeated occurrences matter.

### Steps

1. Freeze the expected population and the matching target scope.
2. Produce separate lists of missing and unexpected identities.
3. Compare occurrence counts or duplicate-sensitive rows when identity membership alone is insufficient.

### Example

All 80 intended partners arrived, but 12 unintended partners arrived too. A missing-only check would have declared success.

### Check

The reconciliation reports omissions, extras and relevant multiplicity differences, not one reassuring percentage.

### Limits

- A target may legitimately contain records outside this run. Define scope before treating all unmatched target rows as defects.

### Evidence and sources

- supports: EXCEPT returns rows in the first query but not the second, and removes duplicates unless ALL is specified. — RS-5EBA53046715516C. A one-direction difference cannot also identify unexpected target rows; distinct-set comparisons can hide repeated occurrences. (Section 7.4)
- RS-5EBA53046715516C: Combining Queries: UNION, INTERSECT, EXCEPT — https://www.postgresql.org/docs/current/queries-union.html

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/make-every-selected-record-end-in-an-explicit-status-bucket

---

## Give omitted, null and blank different test cases

ID: MHC-D-RESEARCH-0982 · Version: 0.1.0 · Kind: checklist
Source: https://vedokrok.com/knowledge/give-omitted-null-and-blank-different-test-cases

An empty-looking cell can carry a very active instruction.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A mass update may clear existing values when the intended action was to leave them alone.

### Avoid when

- Do not standardize all missing representations into one value before checking their semantics. Different fields in the same interface can behave differently.

### Explanation

Write down how this exact interface represents no change, clear the value and set an empty string. Do not infer those meanings from appearance. JSON Merge Patch, for example, gives null a removal meaning; another interface may use an update flag or a different convention.

### Checklist

- Read the field-level update contract, including flags and defaults.
- Test omitted, null, empty-string and explicit-value inputs against an existing nonempty value.
- Inspect the stored result and any downstream message before approving a bulk run.

### Example

A file omits an email column in one test and includes it empty in another. Only the documented behavior determines whether either clears the address.

### Check

Each input state has an observed, documented target effect.

### Limits

- Do not standardize all missing representations into one value before checking their semantics. Different fields in the same interface can behave differently.

### Evidence and sources

- supports: JSON Merge Patch distinguishes an omitted member from a member set to null: null requests removal of that member. — RS-4FB38C20FE966683. This is one interface contract, not a universal interpretation of empty spreadsheet cells or null values. (Sections 1 and 2)
- RS-4FB38C20FE966683: RFC 7396: JSON Merge Patch — https://www.rfc-editor.org/rfc/rfc7396.html

### Vedokrok editorial review

Test what each empty-looking update will actually do

Omitting a field, removing it and storing an empty value are different instructions. Check the exact interface before a harmless-looking blank becomes a bulk change.

#### Keep the rule at the right level

For an object-member update in JSON Merge Patch, an omitted member is left alone, null removes the member, and an empty string sets an empty string. This is not a rule for every JSON value: a patch that is not an object replaces its target completely. A top-level null patch therefore replaces the document with null.

#### Inspect the stored result

In an isolated test record with a nonempty reference field, try omission, null, empty string and an explicit replacement separately. Record both the stored value and whether the field still exists. Test arrays separately if present; Merge Patch replaces them rather than editing selected elements. Check any downstream interpretation before applying the mapping to a real batch.

**Editorial takeaway:** Use the documented field and document rules, then verify the result. If the system needs a stored null to mean something distinct from removal, do not assume an object-member Merge Patch can express that change.

#### Editorial evidence

- supports: JSON Merge Patch distinguishes omitted object members, null-valued members and non-object patches. — RS-4FB38C20FE966683. The member-removal rule is scoped to object processing; the independent reference-field tests are an editorial application. (Section 2: MergePatch pseudocode and following paragraph; Appendix A)
- limits: Merge Patch does not support partial array edits and is not suitable for every JSON structure, particularly when explicit null object-member values matter. — RS-4FB38C20FE966683. Do not transfer these semantics to ordinary JSON storage, JSON Patch, spreadsheets or other import interfaces. Test the actual accepted media type and contract. (Section 1, final paragraph; Section 2; Appendix A, whole-target null example)
- RS-4FB38C20FE966683: RFC 7396: JSON Merge Patch — https://www.rfc-editor.org/rfc/rfc7396.html — Paul Hoffman, James M. Snell

Editorial version: 0.1.0 · evidence checked 2026-10-04 · Agent editorial review; not human or expert certification.

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/find-the-first-boundary-where-the-record-becomes-wrong

---

## Keep the unit attached when moving the number

ID: MHC-D-RESEARCH-0983 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/keep-the-unit-attached-when-moving-the-number

The number 500 is impressively precise about almost nothing on its own.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A value crosses a system boundary, report or manual calculation.

### Avoid when

- Currencies, temperature scales and business-specific units may need context beyond a simple multiplier. Do not improvise safety-critical conversions.

### Explanation

Carry the unit and the quantity definition with the value. Convert only after confirming that both sides measure the same kind of quantity. A correct conversion cannot repair a comparison between different business definitions.

### Steps

1. Record value, unit and what was measured together.
2. Name the target unit and the documented conversion rule.
3. Check one known example and preserve the original value for reconciliation.

### Example

A duration of 500 milliseconds becomes 0.5 seconds, not 500 seconds. Separately confirm whether both systems include waiting time.

### Check

The receiver can reconstruct the value's meaning and distinguish conversion from a changed measurement definition.

### Limits

- Currencies, temperature scales and business-specific units may need context beyond a simple multiplier. Do not improvise safety-critical conversions.

### Evidence and sources

- supports: UCUM distinguishes unit symbols, prefixes and semantic equivalence so numeric quantities can be interpreted with their units. — RS-0C3759023DCB78E8. Compatible physical dimensions do not establish matching business definitions, measurement conditions or conversion dates. (Unit terms and semantic rules)
- RS-0C3759023DCB78E8: The Unified Code for Units of Measure — https://ucum.org/ucum

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/find-the-first-boundary-where-the-record-becomes-wrong

---

## Make every selected record end in an explicit status bucket

ID: MHC-D-RESEARCH-0984 · Version: 0.1.0 · Kind: template
Source: https://vedokrok.com/knowledge/make-every-selected-record-end-in-an-explicit-status-bucket

A total should account for the awkward records too.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A batch summary lists successes but leaves it unclear what happened to the rest.

### Avoid when

- Retries and duplicate inputs require a defined counting unit. Do not mix business identities, attempts and message counts in one equation.

### Explanation

Freeze the selected input population and assign each identity to one mutually exclusive status at a stated cutoff. Include pending and unaccounted states instead of hiding them inside success. This is population accounting; successful accounting is not proof that the resulting field values are correct.

### Template

Run and cutoff: [snapshot]. Selected identities: [total]. Applied: [applied]. Rejected: [rejected]. Pending: [pending]. Unaccounted: [unaccounted]. Next action for unresolved identities: [action].

### Example

A 200-record run has 183 applied, 9 rejected and 8 pending. It is fully accounted for, but it is not 200 successful updates.

### Check

The disjoint buckets reconcile to the selected population, and unresolved IDs are retrievable.

### Limits

- Retries and duplicate inputs require a defined counting unit. Do not mix business identities, attempts and message counts in one equation.

### Evidence and sources

- supports: AWS DMS distinguishes validation-pending, failed, suspended and validated records rather than treating every transferred record as validated. — RS-F1423B1DF36F3A34. A population-accounting ledger can expose unaccounted records but cannot replace comparison of their actual values. (Validation states)
- RS-F1423B1DF36F3A34: AWS DMS data validation — https://docs.aws.amazon.com/dms/latest/userguide/CHAP_Validating.html

No review details supplied.

---

## Import identifiers as text before a spreadsheet can reinterpret them

ID: MHC-D-RESEARCH-0985 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/import-identifiers-as-text-before-a-spreadsheet-can-reinterpret-them

A customer code is a name wearing digits, not necessarily a number.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Identifiers contain leading zeros, long digit strings or number-like codes.

### Avoid when

- Formatting a damaged value as text is too late. Reimport from an intact source rather than guessing lost digits or zeros.

### Explanation

Set the import column to text before values enter a numeric representation. Excel can remove leading zeros or lose precision in long numeric strings. Changing the display afterward does not recover information that has already been discarded.

### Steps

1. Use a controlled import route and declare identifier columns as text.
2. Include test values with leading zeros and more than 15 digits.
3. Compare the imported and exported strings with the untouched source.

### Example

The code 000742 must remain six characters when exported for a matching job; displaying 742 with padding is not automatically the same preservation route.

### Check

A round trip preserves the exact identifier strings, including length and final digits.

### Limits

- Formatting a damaged value as text is too late. Reimport from an intact source rather than guessing lost digits or zeros.

### Evidence and sources

- supports: Excel can remove leading zeros and lose precision beyond 15 significant numeric digits; formatting afterward does not restore already lost information. — RS-9E2FDA3F93273862. Preventing conversion must happen before the affected values enter a numeric representation. (Import as text; long numbers and custom formatting)
- RS-9E2FDA3F93273862: Keeping leading zeros and large numbers — https://support.microsoft.com/en-us/excel/keeping-leading-zeros-and-large-numbers

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/match-the-identifier-together-with-its-issuing-namespace

---

## Choose the clock before assigning an event to a period

ID: MHC-D-RESEARCH-0986 · Version: 0.1.0 · Kind: question
Source: https://vedokrok.com/knowledge/choose-the-clock-before-assigning-an-event-to-a-period

Arrived today and happened today are different statements.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Late messages or delayed processing move events into the wrong reporting day.

### Avoid when

- Source clocks and timestamps can be wrong. A streaming watermark estimates progress; it does not make late arrival impossible.

### Explanation

Keep occurrence time and processing time separate. Decide which clock answers the business question, and define what happens when an event arrives after a reporting cutoff. Ordering records by arrival cannot reconstruct the event sequence by itself.

### Question

Does this report concern when work happened or when the system received it? · Which time zone and cutoff define the period? · How will late events correct or annotate an already issued result?

### Example

A service event occurs before midnight but reaches reporting after midnight. Its business date follows the agreed occurrence-time rule, not an accidental queue delay.

### Check

A deliberately late test event lands in the expected period or follows the documented correction route.

### Limits

- Source clocks and timestamps can be wrong. A streaming watermark estimates progress; it does not make late arrival impossible.

### Evidence and sources

- supports: Event time records when an event occurred, while processing time records when a system processes it; arrival order need not follow event order. — RS-4DB447BFC603BBA8. Choosing the reporting clock and late-data correction policy remains a domain decision. (Section 8.4)
- RS-4DB447BFC603BBA8: Apache Beam Programming Guide — https://beam.apache.org/documentation/programming-guide/

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/join-historical-work-to-the-version-that-was-valid-then

---

## Join historical work to the version that was valid then

ID: MHC-D-RESEARCH-0987 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/join-historical-work-to-the-version-that-was-valid-then

Today's organization chart did not manage last year's sale.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A current master-data value silently changes the interpretation of old transactions.

### Avoid when

- Corrections and retroactive changes need agreed rules. Keeping versions is insufficient when their validity dates or source meaning are wrong.

### Explanation

When the question is historical, preserve or obtain effective-dated versions and match each event to the version valid at that time. A Type 2 slowly changing dimension is one implementation. Joining everything to the current row answers a different question.

### Steps

1. State whether the report needs current classification or historical classification.
2. Use the relevant effective-time intervals or version keys.
3. Check gaps, overlapping intervals and changes that occur exactly at a boundary.

### Example

A customer moves to a new region in July. June sales keep their historical region when the report asks who served them at the time.

### Check

A before-and-after test around one known change produces the intended historical attribution.

### Limits

- Corrections and retroactive changes need agreed rules. Keeping versions is insufficient when their validity dates or source meaning are wrong.

### Evidence and sources

- supports: A Type 2 slowly changing dimension preserves versions with validity dates so historical facts can be associated with the relevant version. — RS-23CF75ED5D468246. Overlapping validity intervals, late corrections and inaccurate effective dates require explicit handling. (Slowly changing dimensions: Type 2)
- RS-23CF75ED5D468246: Understand star schema and the importance for Power BI — https://learn.microsoft.com/en-us/power-bi/guidance/star-schema

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/name-what-one-row-means-before-joining-two-tables

---

## Match the identifier together with its issuing namespace

ID: MHC-D-RESEARCH-0988 · Version: 0.1.0 · Kind: template
Source: https://vedokrok.com/knowledge/match-the-identifier-together-with-its-issuing-namespace

Record 123 in one system has not introduced itself to record 123 in another.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Two systems use the same-looking number for potentially different entities.

### Avoid when

- Identifiers may be merged or reassigned. Record validity periods and ambiguous matches instead of forcing a one-to-one mapping that the domain does not support.

### Explanation

Treat an identifier as a value inside an issuing namespace. Use an authoritative crosswalk when matching across namespaces, and preserve both original identities. A shared name or number is a candidate match, not sufficient proof of equivalence.

### Template

Source namespace: [source system]. Source value: [source id]. Target namespace: [target system]. Target value: [target id]. Mapping authority and validity: [basis].

### Example

Customer 123 in a legacy system maps to partner 900123 in the new system. An unrelated target customer numbered 123 must not capture the relationship.

### Check

Every accepted cross-system match has a namespace-aware rule or traceable mapping decision.

### Limits

- Identifiers may be merged or reassigned. Record validity periods and ambiguous matches instead of forcing a one-to-one mapping that the domain does not support.

### Evidence and sources

- supports: FHIR Identifier separates the namespace in system from the identifier string in value, avoiding reliance on the value alone for identity. — RS-F863D183EF6C607D. A crosswalk still needs an authoritative mapping and rules for reassignment, merges and validity periods. (Identifier.system and Identifier.value)
- RS-F863D183EF6C607D: FHIR datatypes: Identifier — https://hl7.org/fhir/datatypes.html#Identifier

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/name-what-one-row-means-before-joining-two-tables

---

## Follow an accepted job to its actual completion state

ID: MHC-D-RESEARCH-0989 · Version: 0.1.0 · Kind: checklist
Source: https://vedokrok.com/knowledge/follow-an-accepted-job-to-its-actual-completion-state

Accepted is a place in the queue, not a certificate of completion.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- An interface reports that a request was accepted, while the requested change happens later.

### Avoid when

- Do not poll without bounds or treat disappearance of a job as success. Some systems retain results only briefly, so capture evidence through the supported route.

### Explanation

Use the operation identifier and documented status route to follow asynchronous work. Separate acceptance, processing and terminal outcomes in the report. Then check the business result that matters; an interface can finish while a downstream dependency is still unresolved.

### Checklist

- Save the job identifier and the receiver's status-check route.
- Follow the documented polling or notification policy, including failure and expiry states.
- Confirm the relevant destination result before marking the business task complete.

### Example

An upload returns an accepted response. The task remains in progress until its status and destination checks confirm the requested records were applied.

### Check

The completion message cites a terminal operation result and the required business check.

### Limits

- Do not poll without bounds or treat disappearance of a job as success. Some systems retain results only briefly, so capture evidence through the supported route.

### Evidence and sources

- supports: The asynchronous request-reply pattern can acknowledge acceptance with HTTP 202 while exposing a separate route for checking later completion. — RS-2859F99043627211. The operation's documented completion state may still differ from downstream business readiness. (Solution)
- RS-2859F99043627211: Asynchronous Request-Reply pattern — https://learn.microsoft.com/en-us/azure/architecture/patterns/asynchronous-request-reply

No review details supplied.
Related (compare_with): https://vedokrok.com/knowledge/treat-a-timeout-as-an-unknown-outcome-before-repeating-the-action
Related (useful_with): https://vedokrok.com/knowledge/make-every-selected-record-end-in-an-explicit-status-bucket
