Dataset Change Checker

Compare two flat JSON snapshots or two selected Apify datasets. Get added, removed and changed records, with exact old/new field values and checks for ambiguous inputs.

US$0.25 per successfully completed comparison. Invalid or incomplete comparisons do not trigger the product event.

Use this when a catalogue, inventory or listing workflow needs a repeatable comparison without maintaining its own matching and export code. If your existing script already handles these checks reliably, this tool may not add enough value for you.

First comparison

In the Actor's JSON input editor, use:

{
  "before": [
    { "sku": "A-10", "price": 12, "stock": 5 },
    { "sku": "B-20", "price": 25, "stock": 2 }
  ],
  "after": [
    { "sku": "A-10", "price": 14, "stock": 5 },
    { "sku": "C-30", "price": 30, "stock": 1 }
  ],
  "keyField": "sku"
}

Run with 512 MiB memory, a 60-second timeout and a maximum product charge of US$0.25. Wait for SUCCEEDED, then open JSON summary and changes, CSV change records, or the compact Completed-comparison receipt in the Output tab.

The example produces one added record (C-30), one removed record (B-20) and one changed record (A-10, price 12 to 14). A changed field includes presence flags:

{
  "kind": "changed",
  "id": "A-10",
  "fields": [
    {
      "field": "price",
      "before": { "present": true, "value": 12 },
      "after": { "present": true, "value": 14 }
    }
  ]
}

Workflow 1: compare existing Apify datasets

Select the old dataset under Old dataset and the new one under New dataset. Leave both inline-record properties absent. Set keyField to the stable identifier present in both snapshots. The Actor requests READ access only to the selected source datasets and never modifies them.

Your upstream jobs must finish writing before this comparison starts. Save their explicit dataset IDs; this tool does not guess which run is "yesterday" or automatically scrape a replacement. If either dataset changes while being read, the comparison fails instead of returning a partial result.

Workflow 2: compare snapshots from another automation

Send before, after and keyField as JSON through the Apify API. Optionally send compareFields, for example ["price", "stock"], to ignore unrelated timestamps or description changes. Omit that property to compare all fields. Do not send an empty list.

Start one run and retain its run ID. Poll that run's status; do not start a new paid run merely because your HTTP client timed out. After SUCCEEDED, use the returned defaultKeyValueStoreId to retrieve RESULT and CHANGES.csv. A failed run may have partial files; file existence alone is not proof of successful billing or completion. The standard API is described in Apify's run documentation.

Comparison rules

• Record order does not matter. Matching is by one unique string or finite-number identifier.
• Identifiers must be present, non-null and nonempty; duplicates are errors. Numeric 1 and string "1" are different identifiers.
• Flat JSON values only: strings, finite numbers, booleans and null. Nested objects and arrays are rejected. Comparisons are exact and type-sensitive, with no rounding, trimming, fuzzy matching or coercion.
• Missing values differ from null. A missing value is represented by {"present": false}. A field absent from both complete snapshots is rejected if explicitly selected, helping catch misspellings.
• Changing an identifier is an addition plus a removal, not an update to the same record.
• At most 10,000 records across both snapshots, 5 MiB combined input and 8 MiB per output artifact. A comparison whose expanded output exceeds the cap is rejected.

Warnings are not deletion instructions

In Apify Console, use beforeEmpty: true or afterEmpty: true to explicitly select an empty snapshot, with that side's records and dataset ID absent. Console can omit optional empty arrays, even from its JSON editor. An omitted source is always an error, never implicitly an empty snapshot. Direct JSON API inputs can still use before: [] or after: []. Do not combine an empty flag with another source on the same side.

The tool warns if a previously populated snapshot becomes empty or its count falls by more than 50%. These warnings indicate possible incompleteness, not proof of an upstream failure. They do not validate business accuracy or authorize deleting downstream records. Inspect flagged runs before applying their changes. This Actor performs no downstream actions.

Price

One US$0.25 comparison-completed event after the JSON result and CSV are both saved. The event is attached to one compact default-dataset receipt so the completed output is visible and compatible with Apify's automated Store checks. A successful comparison with no changes is still one comparison. Invalid or incomplete comparisons produce no receipt and do not trigger this product event. There is no per-record or start charge, and platform usage is not passed through separately.

Your own source storage, retention, downloads or other platform services may still have charges under your Apify plan. Check the listing's actual pricing before starting a run. This Actor does not offer a refund policy beyond the applicable platform terms.

Results, privacy and deletion

RESULT is the canonical JSON output. CHANGES.csv is a human-readable export with formula-prefix hardening, which can alter display text. Use JSON for lossless machine processing. The default dataset contains one compact receipt with counts, warnings and the two artifact names; it does not duplicate the source records or full change payload. CHECKPOINT holds progress metadata and an input fingerprint, not an extra copy of source records.

Input and results use Apify run storage; this tool has no separate database or payload telemetry. Application error messages use fixed codes rather than record values. Each limited-permission run executes under the customer's account and can access only its own storage plus resources explicitly selected in the input. The code does not copy run data to the developer.

Apify account settings can nevertheless let a customer opt in to sharing public-Actor runs with developers, and including a run URL in an Actor issue shares that run, its input, logs and default storage with the developer for support. Review the account's Share run data with developers setting and never attach a sensitive run to an issue. Platform processing, access controls and terms still apply.

Retention depends on the customer's plan. Apify currently retains a Free account's ten most recent runs and their storage for four months; unnamed runs beyond the most recent ten expire under the account retention rules. Paid accounts follow their configured retention, while named stores are retained indefinitely. Download needed results before expiry and inspect the current account settings; this Actor makes no fixed deletion-time promise. Use the key-value store's record/store deletion controls when appropriate. See Apify retention rules and storage deletion controls. Do not submit credentials, payment details or sensitive personal records as comparison data.

Errors and support

Code	Action
DUPLICATE_IDENTIFIER, MISSING_IDENTIFIER, INVALID_IDENTIFIER	Fix the key or upstream records; no automatic deduplication is performed.
EXACTLY_ONE_SOURCE_PER_SIDE_REQUIRED	Supply inline rows or a selected dataset for each side, not both.
UNKNOWN_COMPARE_FIELD	Correct the field name or omit compareFields.
FLAT_JSON_RECORDS_REQUIRED	Flatten nested data upstream.
SOURCE_READ_FAILED, INCOMPLETE_SOURCE_READ, SOURCE_CHANGED_DURING_READ	Check access and that the upstream snapshot finished successfully.
RECORD_LIMIT_EXCEEDED, INPUT_SIZE_LIMIT_EXCEEDED, OUTPUT_SIZE_LIMIT_EXCEEDED	Reduce input or selected fields; do not split snapshots in a way that changes matching semantics.
INVALID_CUSTOMER_BUDGET, INSUFFICIENT_CUSTOMER_BUDGET	Set an explicit finite product budget of at least US$0.25 for a paid comparison.
BILLING_RECONCILIATION_REQUIRED, SAVED_RESULT_UNAVAILABLE	Inspect the original run's charges and contact support before retrying.

Use the Actor's Issues tab for support with the error code and a small synthetic reproduction. Never post API tokens, private datasets or financial information. Custom development, live onboarding, scraping, standalone CSV uploads and always-on monitoring are not included.