Load history into a data series

A new series rarely starts from a blank page: usually there is an Excel tab going back years that you want to bring in. Load history does exactly that — it loads the rows of a CSV or XLSX file into the series in an audited way: first a dry run (a write-free preview), then a commit. A load can be reverted later.

It differs from live recording (see Record into a data series) in being bulk and historical: the sample time is read from the file’s Date (and optionally Time) column, in the site’s time zone — the recording surface’s backdating limit is not a barrier here.

Before you start

• Data-import permission: data:import to load and dry-run; the stricter data:import:commit permission to commit and revert. Without the permission the buttons are shown disabled, with a reason.
• The paid add-in switched on (the same one the data series need).
• The series already has its columns — you map the file’s columns to these. If not, first set up the series.
• Where: on the Management → Data series page, the Load history button in the series’ row.

The steps

1. Open the Load history page for the series and upload a CSV or XLSX file. The file itself is not stored — only its raw rows go to a temporary workspace.
2. Map the columns. Mark which one is the Date column (required) and — if separate — the Time column (optional; empty means local 00:00). Then map each value column to the matching series column; unmapped columns are left out (they do not become zeros). Save the mapping.
3. Run the dry run. The system measures every row without writing and sorts them into four groups: OK, Invalid (type/limit), Bad time (invalid or future sample time), and Duplicate (a time already in the series). Each failing row shows a reason.
4. Commit the load. The valid (OK) rows are loaded into the series. Duplicates and invalid rows are left out — nothing is overwritten.
5. Revert if needed. A committed load can be undone with Revert load: the loaded samples are deleted from the series, while values recorded in the app stay untouched.

The dry run is a forecast, not a contract

The dry run is true for the rows measured now. The final classification — duplicates included — is re-checked at commit; if a column’s limit changed in the meantime, the commit stops so you can review again.

Which Date/Time formats are recognised?

The Date column can be a real Excel date cell, YYYY-MM-DD (or YYYY.MM.DD) text, or an Excel date serial. The Time column can be a real Excel time cell, HH:MM (or HH:MM:SS) text, or a number-formatted time. If a single column holds both the date and the time, just select it as the Date column. Anything we cannot interpret (for example “morning”) is reported as Bad time — we never silently assume midnight.

Duplicate = the same sample time

Two rows are a duplicate when their sample time (and, if the series uses batches, their group key) matches an existing row. Re-loading therefore does not duplicate data: the second occurrence is left out. A future sample time is never accepted — you cannot take a sample ahead of time.

A corrected sample keeps its slot

If you corrected a previously loaded sample through the UI (creating a new version of it) and then load the same file again, that row is reported as a duplicate — the load neither overwrites it nor resurrects the pre-correction value. This is deliberate: a manual correction outranks a bulk load. If you do want to load it again, revert the earlier load first.