Overview
The purpose of this article is to explain how the platform creates a History sets.
If someone dives into the sys_history_line and sys_history_set tables without knowing the details, what they see can be confusing.
For a detailed explanation of the difference between Audit and History sets be sure to check the official documentation here.
History Sets
To avoid generating a new history list from sys_audit data every time someone views a record history --this would be time consuming-- the platform uses a concept called 'History sets'.
These can be considered personalized historical views as they are generated based on user profile information.
The sets are then saved to the sys_history_set and sys_history_line tables.
So these tables must be huge, what about performance/space concerns?
As the end goal is to avoid querying the large sys_audit table the system limits History Set and History records by:
- Having the table cleaner delete History Set records that have not been updated in 30 days.
- Using table rotation to rotate between four History tables every seven days. This means the system drops History records that are older than 28 days.
How does the System know when to create a new History Set?
A new history set is created for each possible combination of the following user profile factors:
- Language
- Timezone
- Domain
- Date/Time Format
If the date of the last update on the record and the last entry on the sys_history_set table differ then that will also trigger the creation of a new history set.
Example
Let's take 3 users:
Mr White:
- English Language,
- IST Timezone,
- Global Domain,
- dd-MM-yyyy HH:mm:ss Date/Time Format
Mr Pink:
- English Language,
- GMT Timezone,
- Global Domain,
- dd/MM/yyyy HH:mm:ss Date/Time Format
Mr Brown:
- English Language,
- GMT Timezone,
- Global Domain,
- dd/MM/yyyy HH:mm:ss Date/Time Format
When Mr White tries to view the historical activity of a record a new history set will be generated according to their profile settings, the same will happen with Mr Pink.
When Mr Brown tries to view the historical data, the platform recovers the history set created for Mr Pink instead of generating a new one thus avoiding an expensive query on the sys_audit table.
Why can this be confusing?
Let's take a look at the sys_history_line table:
- As you can see from above the same row is essentially repeated three times and each row is associated to the same History Set.
- The Update time and User name are identical
So what's going on here?
Each row corresponds to a different history set permutation although all are grouped under the same history set ID.
The update time and user name correspond to the sys_created_on and the sys_created_by fields of the record itself: