Compare
Added in v0.4.0
Is TEST the same as PROD? What changed in the item master since last week? Did the load land the way the spreadsheet said?
Compare answers those questions in a few clicks. Start from a result in the Console, pick up to three other results of the same endpoint, and Dobermann lines up their records and shows you, cell by cell, what is different, what is missing and what is the same.
- Environments side by side. DEV, SIT, UAT and PROD, from the same GET, in one grid.
- One environment over time. This morning’s run against last week’s: what moved.
- Before and after a load. Pull the data, run your batch, pull it again, compare.
Compare reads results you already have in History. Nothing is sent: no request goes to any API, so you can compare production data as often as you like.
Compare is in Beta. It works, it’s fast at any size, and it’s still settling. Tell us what you find on GitHub: what’s confusing, what’s missing, what’s wrong.
Start a compare
A compare always starts from a result.
- Open a finished run’s Console and go to the Completed tab.
- Choose the view whose columns you want compared. Compare reads every result through this view, so pick or edit it first.
- Click Compare, beside Chain. The Compare tab opens on its first step, Select.
- Choose Other results from this endpoint, then tick up to three results to compare.
- Click Next.
The result you started from is the Baseline, BL: everything is measured against it. Pick another in the Baseline column if you’d rather. The others are C1, C2 and C3, in the order you ticked them. The four slots across the top fill as you pick, each result in its own colour, and stay there on every step.
Compare reads the results, then opens Match. The step bar at the bottom shows how far it has got, on the step it leads to. Back and Next, beside it, move between the steps.
- A run’s environment shows beside each result, with any header values
it ran with (
DEV 880,DEV 840), so one environment run for two companies is two results you can tell apart. - A result marked Incomplete didn’t fetch every page: its slot is striped. The records on the pages it missed will show as missing.
- Compare is on the Completed tab only, and comes on once the run has ended.
- With only one result for the endpoint there is nothing to compare: run it again, or run it in another environment.
- Other results from any endpoint and A file from your computer are coming soon.
Match
Match decides which record in one result is the same record in another. It opens as soon as you click Next, while the results are read; the step bar shows how far that has got.
Under Match on, on the right, tick the columns that identify a record: a business key such as
sku or itemCode, or a few columns together. Each column shows a sample value from BL. A database
id usually won’t work: each environment numbers its records for itself. Compare checks the match as
you tick.
How BL matches C1 (or each result, with more than one), on the left, shows what the key does over every record. It lines up under the slots: each row’s picture sits under BL (hover it for its name), and each result’s count sits under that result’s slot.
| One to one | Records that pair up. These are compared |
| Repeated in C1 | One to many: BL has a value once (or not at all) and C1 has it on several records. These are compared too: BL’s record against each of C1’s |
| Only in BL | In the Baseline and not the other. They show as Missing C1 in Results |
| Only in C1 | In the other and not the Baseline. They show as Missing BL in Results |
The Baseline is always the “one”. A value on more than one of BL’s records stops the compare, in a box headed Match more than one record: can’t be compared: Repeated in BL and Repeated in both. Tick another column so BL holds each value once, or make the other result the Baseline.
In the repeated rows a cell shows BL’s records, the picture, then that result’s: 36 ●< 200 is 36
records in BL and 200 in C1 on the same values. Large counts are shortened (105K); hover the cell
for the exact numbers and how many values they share.
When some values repeat but nothing stops the compare, the verdict says so in amber: BL’s record is compared with each of them. Go on to Columns.
See the repeated records
Click a count in a repeated row to see those records:
- One row per value, the “one” side’s record: BL’s for Repeated in C1. The toggle at its start (2 in C1) opens the other side’s records for that value alone, in that result’s colour.
- Repeated in both lists the values; opening one shows both sides’ records.
- Values the “one” side doesn’t have are missing as well as repeated, so they’re left out. Tick Show the … values BL doesn’t have to list them too.
- Open BL in the Console and Open C1 in the Console show a result in full, every field included.
- ‹ and › page through the values, 25 at a time.
When one value is on thousands of records (a user id on every record, say), the window says why that can’t be a key and lists the values with how many records each side holds.
The slots’ menu
From Match on, click a slot for its menu: Make Baseline (it swaps places with BL, and the key is checked again), Remove from the compare (the results after it move up a place), and Open in the Console. On Results the menu also has Hide and Show.
Back returns to Select, to pick a different Baseline or other results.
Columns
Columns lists the view’s columns: the ones to compare, and the ones only to show. For each one:
- Show puts the column in Results.
- Compare says how its values are compared:
| Strict | The default. The value and its JSON type must be the same: "00123" is not 123, and "12.50" is not "12.5". Nothing is reformatted, because a dropped leading zero is exactly the kind of change you’re looking for |
| Relaxed | Spaces at either end, repeated spaces and letter case don’t count; a number equals the same text (12.5 and "12.5"); blank, null and missing are one value. Leading zeros still count |
| Don’t compare | The column shows in Results, but a difference in it is never a DIFF. Use it for ids and timestamps you want to see beside each record |
The key columns are always compared strictly.
Which rows? chooses the rows Results starts with:
- Everything: every record, matched or not.
- All of BL: every Baseline record, with what the others have of it.
- All of C1: every record of the other result.
- Matching only: records found on both sides.
It only sets the starting filter. Every row is still counted, and the Status filter in Results changes it at any time.
Click Generate to build Results.
Results
One grid, one row per record.
The Status column says what Compare found:
| Same | Every compared column matches |
| DIFF | At least one compared column differs. With three or four results, the letters say which: DIFF BL means the others agree and the Baseline is the odd one out; DIFF C2 means C2 differs from the Baseline |
| Missing | A result doesn’t have the record: Missing C1, Missing BL |
A column’s heading shows a count of the rows that differ in it. A cell that differs is marked in its result’s colour.
A value repeated in a result (one to many) is a row for each of that result’s records, each compared
with BL’s one record, so one can be Same and another DIFF. The other results’ records for
the value repeat on every one of those rows. Add id as a shown column if you need to tell the rows
apart.
Relationship, between the key and Status, shows how each row’s key meets the results, as Match’s pictures: one to one, one to many, not in every result, not in BL. Hover a picture for its name. It appears when the rows don’t all have the same relationship.
Across the top, beside the slots (under them, in a narrow window):
- The slots, BL first. Click C1, C2 or C3 for its menu: Hide (it isn’t compared while hidden) or Show, Make Baseline (it swaps places with BL), and Open in the Console. Click BL to make another result the Baseline, or to open it. A striped slot didn’t fetch every page; its menu says so. Hiding a result that repeats values folds its extra rows away, as if it weren’t in the compare; Show brings them back.
- Side by side gives each column one sub-column per result. Compact gives each column one cell, with the values that differ in pills. Hover over the headings to see which result is which.
- All columns or Diffs. Diffs hides the compared columns that no row differs in, so a wide record shows only what changed. It only hides them from view: they are still compared, and one comes back as soon as a row differs in it. The key and the columns you set to Don’t compare always show.
- clears every filter. It’s lit while a filter is on.
- Columns: tick the columns in the compare. A column you set to Don’t compare is marked show only. Unticking a column removes it from the compare, so the counts change with it.
- Search finds rows, with the same syntax as the Console’s search,
@columnterms included:@status=active,@quantityOnHand>10. Type@to pick a column. Status and Relationship aren’t searched: their own filters do that. - The count of rows shown.
- Export saves to Excel (see Export).
Click a column heading to sort.
Filter the rows
- Status: click the arrow beside Status. Each line has a tick box under Show, to add or remove it, and a radio under Only, to see that line alone: Only on DIFF shows just the differences. Click Only again on the line that’s alone to see every line again. The counts say how many rows each line has.
- Relationship: click the arrow beside its picture. It works as Status does: Show, Only, and how many rows each relationship has. Only on one to many shows just the repeated rows.
- A column’s count: click the count in a column’s heading to show only the rows that differ there. Click several and you see the rows that differ in all of them.
A filter that’s on shows a small beside it. Click it to clear that filter alone, without opening its menu.
A record in full
Click a row, or right-click anywhere on it, to open that record from every result:
- Table lists the view’s columns, one result per column, with the differences marked.
- JSON shows each record exactly as its system sent it, side by side, lines aligned.
- DIFF: All marks differences in the columns you aren’t comparing too, more faintly. Selected marks only the compared ones.
- fills the window, for wide records.
A long value, or one holding JSON, is cut short in the grid. Click it to open that value in full, formatted, as in the Console.
Export
Export saves what Results shows, as an Excel workbook: a Summary sheet (the results compared, the key, the columns and the counts), then the Details, one row per record shown, with the filters and search applied. Past Excel’s row limit, the rows run on into Details 2, Details 3 and so on.
Values keep their type: text stays text, leading zeros and all, and numbers stay numbers.
Saved compares
A compare is saved once Next on Select picks its results, then as you build it: the results, the key, the columns, Strict or Relaxed, which rows, and the columns you hid. Change its results (add C3, say) and Columns and Results wait for Match again, even after you close and reopen it, until it has been compared with exactly those results.
In the Hub, Reporting lists them. Click one to open it on Select, with everything as saved. Nothing is read until you move on: pick other results if you like, or click Results in the step bar to go straight there. Compare reads, matches and compares the results on the way, and the steps before Results tick as it goes. Right-click a compare for:
- Open.
- Duplicate: a copy to change.
- Delete: removes the compare only. The results stay in History.
Related Topics
- Console: where a compare starts
- Named Views: the columns a compare reads
- Viewing Results: History and the saved results