A reformatted file makes a line-based JSON diff useless: every line changes and the one real edit hides in the noise. CodePrettify parses both sides and compares values — key order and formatting ignored, array items matched by id — with two editable inputs and everything processed on your device.
The two configurations below differ in key order, indentation, and one actual value. A line diff flags nearly every line; a semantic JSON diff reports exactly what changed.
The version currently running, pretty-printed with keys in their original order.
{
"service": "checkout",
"replicas": 2,
"rollout": {
"strategy": "RollingUpdate",
"maxUnavailable": 1
},
"regions": ["eu-north-1", "eu-west-1"]
}Reordered keys and different formatting — but the only value change is replicas.
{
"rollout": { "maxUnavailable": 1,
"strategy": "RollingUpdate" },
"regions": ["eu-north-1", "eu-west-1"],
"replicas": 3,
"service": "checkout"
}Comparison is listed among the operations that never leave the extension or the Windows app. Unlike compare-JSON-online tools, nothing you paste is uploaded.
Use the two configurations above, or any pair of payloads you are reviewing. The important habit is checking which side is which before you read added and removed markers.
Press Ctrl+Alt+D, or choose More Actions → General tools → Compare. The Command Palette (Ctrl+Shift+P) finds it by name as well.
The left input starts with the current document and the right input starts empty. Each side has its own Use current document and Clear controls, so you can rebuild either input at any time.
Both sides are fully editable. Semantic availability is detected from the two inputs, not from the open file — you can paste two JSON payloads while viewing a JavaScript or CSS document and still get a structured comparison.
Before comparing, Swap sides exchanges the actual inputs. Direction matters: an entry reported as added on one orientation is reported as removed on the other.
Semantic mode is selected automatically when both inputs parse as JSON, JSONC, JSON Lines, YAML, or TOML — the mode toggle appears only when semantic comparison is possible. Ctrl+Enter runs the open dialog’s primary action.
For the sample, expect a single changed value: replicas from 2 to 3. Change sets are capped at 5,000 entries, and the counters show a trailing ellipsis when truncated. Identical sources show a centered no-differences message instead of an empty list.
After a comparison, Swap Sides flips only which side is displayed; use Edit inputs to change the actual text and compare again. The Windows app adds Previous/Next difference buttons for stepping through a long result.
A comparison is evidence only when you know how it was computed. Confirm these five points.
Semantic mode compares parsed values. Text mode compares exact lines with synchronized pane scrolling, and in the Windows app adds an Ignore whitespace option that disregards all whitespace inside lines. When either side fails to parse, the comparison falls back to text.
Left starts with the current document; paste, edit, clear, or swap either side.
Both sides parse as JSON, JSONC, JSON Lines, YAML, or TOML — semantic. Otherwise — text.
Flip the displayed sides, return to Edit inputs, or step differences in the Windows app.
In JSON Lines, the app accepts JSONC comments and trailing commas per record; the extension requires strict JSON per line.
Only value and structure differences are reported in Semantic mode.
id, _id, uuid, or guidVery large text comparisons — over roughly 5,000 combined lines in the extension, 20,000 in the app — switch to a faster positional comparison that can overstate changes when lines shift.
Independently pasted or edited text never produces a misleading jump.
Selecting exactly two supported files in Explorer can expose the Compare command directly.
| What you see | What it means | Best next action |
|---|---|---|
| No Semantic/Text toggle | At least one side does not parse as a supported structured format, so only text comparison is possible. | Repair or complete the invalid side, then compare again for a structural result. |
| Change counter ends with an ellipsis | The change set hit the 5,000-entry cap and the list is truncated. | Compare the specific subtree or record range you care about instead of the whole document. |
| Noisy diff on very large text | Above roughly 5,000 combined lines (extension) or 20,000 (app), the faster positional comparison can overstate changes when lines shift. | Compare smaller sections, or use the Windows app for the higher threshold. |
| Indentation-only line changes | The lines differ only in whitespace, not content. | In the Windows app’s Text mode, enable Ignore whitespace; or use Semantic mode when both sides parse. |
| A reordered array reports many changes | The elements lack a shared unique id, _id, uuid, or guid key, so the array is compared positionally. | Keep a stable identifier on both sides to get identity matching. |
| Centered no-differences message | The two sources are identical in the selected mode. | Nothing to review — swap in the next pair with Use current document or Clear. |
In the extension, open Compare, keep the current document on the left, and paste the second file on the right — each side is fully editable. The Windows app additionally lets you browse two files into a dedicated comparison tab, and selecting exactly two supported files in Explorer can offer a Compare command.
Not in Semantic mode. It compares parsed structure, so reordered object keys, indentation, and line-break differences produce no changes. Switch to Text mode when you specifically need exact line-by-line differences.
Both sides are parsed into structured data and compared value by value instead of line by line. Arrays whose elements are all objects with a unique primitive id, _id, uuid, or guid key on both sides are matched by identity, so a reordered array reports real edits instead of rewriting every position; all other arrays are compared positionally.
No. Comparison runs entirely on your device in both the extension and the Windows app. Nothing you paste or open is sent to a server — the practical difference from compare-JSON-online tools.
Paste both versions, let semantic mode strip away the formatting noise, and review the actual changes on your device.