Compare two small CSV files by ID
Fictional practice files from HoldMyTwist

These files contain invented stationery records. Amounts have no currency.
Keep the originals unchanged. Unzip this folder before selecting CSVs.

START WITH old.csv AND new.csv
Both have five data records and the headers ID,Item,Amount.
Open https://rowslice.com/compare-csv/ and select both files.
Check old.csv is the Old file and new.csv is the New file.
Use Swap Old and New if needed.
Read every ID in both previews before comparing. Old IDs: 0001, 0002,
0003, 0004, 0005. New IDs: 0005, 0002, 0001, 0006, 0004.
Each ID is nonblank and appears once within its own file. The leading zeros
are part of the ID. The different order does not change which record it is.
Match rows on ID. Choose 'any other column is different'. Leave both ignore
options off for the exact-text comparison. Press Compare Files.

EXACT ANSWER
Added: 0006, Lamp, 40.00.
Removed: 0003, Folder, 3.75.
Changed: 0002 Amount changes from 12.50 to 13.00.
Changed: 0004 Item changes from 'Pen' to 'Pen ' with one trailing space.
Unchanged: 0001 and 0005.
Counts: 1 added, 1 removed, 2 changed, 2 unchanged, 0 rows left out.
Old total: 1 removed + 2 changed + 2 unchanged = 5.
New total: 1 added + 2 changed + 2 unchanged = 5.
Counts are a consistency check. Also check each ID and changed value.

WHITESPACE CHOICE
If you repeat the same pair with 'Ignore spaces at the start and end of
values' on, 0004 is counted unchanged. The answer becomes 1 added, 1 removed,
1 changed, 3 unchanged. This deliberately ignores that trailing-space change.
RowSlice says this option also affects keys. Do not use it without deciding
whether those differences matter and checking the resulting keys.

PRACTICE THE STOP CHECK
Use duplicate-old.csv with new.csv only to inspect the previews. There are
six old records. The sixth repeats ID 0002. Stop before relying on a comparison.
Use blank-old.csv with new.csv only to inspect the previews. There are six
old records. The sixth has no ID. Stop before relying on a comparison.
For your own exports, resolve missing or repeated IDs against the original
source. Do not invent an ID or delete a legitimate record to silence a warning.
The comparison tool can still return results for these cases. A zero 'Rows
left out' count does not establish that every ID is valid.

SCOPE
This manual check is for small nonsensitive files whose complete records you
can inspect. A preview may show only part of a larger file. If you cannot
account for every record, stop and use a suitable validation process.
The example does not test hidden characters in keys, conflicting duplicates,
large files, malformed CSVs, changed headers or composite keys. It is not a
security or privacy audit, nor a guarantee for other files.
expected.json is a machine-readable answer key; you do not need to run code.
