The example calculates expected sum six beside an illustrative retained five. The first data revision is 2,3; the second is 2,4 while report.qmd stays unchanged. Selecting the changed-data scenario exposes the mismatch. The page performs local arithmetic and record comparison, not a Quarto render, cache refresh or GitHub Pages deployment, so test data freshness in Quarto itself before publication.
Give source and data separate identities
The minimal report reads a named count column in external counts.csv. Record exact document revision, data bytes/hash, engine and environment. A shared filename does not mean the contents are unchanged. Use freeze:auto and cache:false with no additional engine-level or memoized cache for this test. The first expected sum is 2+3=5. In the second case, retain the document but change the data to 2+4=6; that gives a simple question about which input the displayed value represents.
The helper has editable numeric data and a separately labeled illustrative recorded result. It never edits an actual CSV or guesses a hash. A matching number alone cannot identify the source revision either: retain the input/output binding alongside the value. If a native log points to a different path than expected, investigate it rather than altering the expected result to fit the page. Keep intended calculation and observed result in different fields so changing one cannot silently overwrite the other.
Distinguish rendering from execution
Quarto documents freeze:auto for global project rendering based on computational source changes, while changed external inputs need deliberate attention. An unchanged report may reuse a frozen result despite a CSV change. A successful global render is therefore not proof the current data was recomputed. The example's retained-five scenario is a freshness failure against expected six, not a claim that a real render ran or that every Quarto project necessarily reuses the same output.
The documented single-document incremental render executes code despite freeze. Confirm the pinned version and engine before using quarto render report.qmd for the actual test. Keep cache:false and the no-other-cache assumption explicit: invoking an engine does not establish fresh data reading when another computational cache can return an old value. Keep the engine log, new input hash, and output 6 together so the record shows which data the computation read.
Review the bundle after accepting the number
Frozen computational state in _freeze differs from the complete generated website. Quarto's publication workflow can use versioned results in another rendering environment without rerunning the original computation. Inventory generated HTML and required assets separately, then later inspect actual Pages URL, base paths, asset loading and displayed result. A frozen directory is not the published bundle, and a deployed-looking page does not identify the input data that produced its number.
Before a later publication, answer which data produced this value. Preserve source, environment, cache policy, frozen-result identity, output revision and observed run under one traceable case. If source, data, or settings change, repeat the computation and keep each run with its original inputs. Preserve historical evidence without deleting shared caches or source files. The local comparison gives a concrete freshness question and expected result while the native Quarto artifacts and actual execution logs remain authoritative for deciding what is ready to publish.
Make your own working example
Build an original Quarto freshness comparison, explicitly not a renderer, cache manager or deployment tool. Document report.qmd remains unchanged. Initial external counts are 2,3 with expected 5. Changed-data scenario 2,4 expects 6 while an illustrative previously recorded value remains 5.
Provide first/second value fields, illustrative recorded-result field, Compare, Changed external data, Expected fresh value, and Reset. Compute actual arithmetic from finite numeric fields. Compare current expectation against the separately entered sample result. Changed-data button sets 2,4 and retained 5, showing mismatch. Expected fresh value sets sample 6 but labels this illustrative agreement, with actual native execution still pending. Never imply that button executes Quarto.
Show freeze:auto, cache:false, no-other-cache assumption, global freeze reuse versus explicit single-document execution, and a native evidence checklist: source/data hash, engine/environment, logs, frozen-result identity and generated bundle. Validate both trimmed data fields before changing the recorded field. Blank or nonfinite inputs and nonfinite sums hold with a specific message, preserving the recorded value; out-of-range arithmetic is not a freshness mismatch. Editing clears previous comparison. Keep website assets/Pages URL/base-path checks separate, no repository changes, cache deletion, GitHub actions, publication or fabricated log.
Use one self-contained HTML file with no remote dependency. Keep the initial sample, output, and labeled controls near the top. Use at least 16px text, strong contrast, visible focus, and a stacked layout at 360px without horizontal overflow. Use explicit type=button click handlers and explicit DOM selectors; do not depend on native form submission. Clear stale feedback when task inputs change. Render entered text safely. Keep practice data local to the current page; do not imply saving, sending, or production integration. Return only the finished HTML.
Clarify the freshness question
Keep native data and execution evidence with the output. In an HTML review ledger, NOS supports visual revision of eligible static notes without AI regeneration. Clarify which input revision a reviewer should check before the eventual Pages publication.