During the most recent week we have seen a few problems with double-entry validation.

Obiously double entry makes no sense if assumed validation is done, but in practice no validation is taking place.

The three situations are:

1. Do NOT use double entry if the command AUTOSEARCH is used in the field which is used for the identification of records.

2. If the field for double entry is computed based on some other fields and that computed field only occurs after entry of a number of fields, then the validation is also problematic.

3. In a relate situation. When creating the file for double entry only the top level entries are created automatically. Not for the lower level files.

For the situation number two (Created link field) it is quite obvious that the programme cannot validate entry of records before the "linking field" is created. There is no way to find out with which record to compare values in the other file before the "linking" information is created. For situation one (AUTOSEARCH) the situation is that the "pointer" to a given record might change in the autosearch field therefore creating a complicated "link" situation that changes during a given entry.

For the development it is essential that such problems are reported. Following detection they are added to a list of known problems to be solved. It is good practice to verify functionality of any software before doing a lot of work in particular with complicated data structures or situations.

Jens Lauritsen EpiData Association