The relationship feature allows Saada operators to set persistent links between data stored into the SaadaDB. These links can be used to browse data and also to select data by using the matchPattern statement of SaadaQL. Each link of a relationship can be qualified with a set of values which are part of the relationship definition.

 Some utilisations of relations in Saada

• Instrumental links  : The relations can be established between reduced data and instrumental data. For example, we can link sources and flatfield pictures and qualify each link with values showing an instrumental feature (efficiency eff ; pixels pix_x et pix_y...). So, each source will possess a ... on the flatfield picture of the camera which is at the origin of the detection. De cette manière chaque source possèdera un pointeur sur l’image flatfield de la caméra àl’origine de sa détection. Moreover, we will be able to select the sources by setting constraints on the detection parameter.

• • Pattern SaadaQL excluding the sources detected in the rectangle delimited by pixels (12,23,344,356) :
    matchPattern{"RelationName"
    , Qualifier("pix_x, "][", 12,23)
    , Qualifier("pix_y, "][", 344,356) }

• Link between theoretical an experimental data . The relations can be used to link simulated data with theoretical data. A SaadaDB can contain a collection of observed data and can also contain several collections of simulated data which correspond to a known pattern. So we can link simulated and observed data to links qualified with a value showing an agreement between the pattern and the observation (reliability).

• • Pattern SaadaQL screens both the entries discribed by the first pattern (whose reliability is higher than 99%)and the entries discribed by the second pattern (whose reliability is higher than 80%) :
    matchPattern{"RelationWithModel1"
    , Qualifier("reliability, ">", 99,0) }
matchPattern{"RelationWithModel2"
    , Qualifier("reliability, ">", 80,0) }

 The examples given below can be implemented if you insert some column whose aim is to contain the qualifier’s value (pixels, efficiency, reliability,...) It’s working as long as we know the number of counterparts. The relations in Saada are working and the number of counterparts does not matter.

• A relation may for example link pictures with all sources detected in the pictures. Such links can be qualified with detection position in pixels coordinates and with the detection reliability. During the picture consultation it is possible to have an immediate access to all sources detected.

>>The relations in Saada can also be used to implement a datamodel. In the SaadaDB the data are stored in different categories(pictures, spectra). A data collection which corresponds to an instance of a datamodel (an observation for example) will be spread out in the base. So, the relations will maintain the links between all the elements of this instance. In this case, the links can be qualified by the identifier of the observation.