From submitted data based on current experience at ArrayExpress[arr] and discussion with curators, most submitted ADF contain many errors and very seldom comply with the ADF specification. It is mainly due to the fact that files are created by hand. So, usually, the item names are incorrect. However, data content is most of the time correct and the submission should be accepted.
In conclusion, the tool is supposed to be flexible enough to allow usual mistakes (see Annexe sec:ADFusualmistakes), even if they are corrected during the checking process. This flexibility allows to have two levels of checking:
And, for convenience, the tool could have during this step two execution modes:
- Relax : allowing the usual mistakes (if they can be identify as well).
If the data are identified, the converting process can be done.
- Strict : file must exactly match the specification.
- A complete mode, which checks whole data;
In that case, the process will not stop if an error is identified.
- A step-by-step mode: once an error is found, the process will stop, allowing a correction of errors one by one (for small data set or known small error numbers);