The Theseus Operator Set

Theseus operators are designed to process data as fast as possible, especially when a set of data is arriving incrementally from a remote network data source. Operators share some common definitions and traits, which are listed below:

• Each data variable processed by operators is a relations.
  
  
• Relations consist of a list of attributes and zero or more tuples of data that contain values corresponding to the attributes of the relation to which they belong.
  
  
• Relations are streamed to/from operators during execution. When streamed, the set of tuples (each with a pointer to the attribute list for the relation) is communicated to the consuming operator from the producing operator. Communication of this set of data is followed by an end-of-stream (EOS) token, to denote that all of the data for that relation has been communicated prior.
  
  
• There is limited support for literal relations in the plan. To express a literal relation, one simply can use double quotes and the "|" character to denote the next tuple. Literal relations always have the same schema ("value char").
  • EXAMPLE: select ("foo|bar", "value != 'foo'" : output) results in only "bar" being emitted.
  
  
• NULL literals (i.e., streams that consist of only an EOS) can be specified by using the reserved keyword NULL.
  • EXAMPLE: select ("foo|bar", NULL : output) results in both "foo" and "bar" being emitted.
  • Note: there has not been extensive testing of NULL in all operators, under all conditions. The most common use of NULL is as a parameter to the Join operator, to indicate that the cartesian product is desired.
  
  
• Several non-relational operators automatically join their input with their output - this is called a dependent join. The Wrapper and Xquery examples demonstrate this feature.
  
  
• Tuples communicated between two operators are not guaranteed to arrive in the order sent.
  
  
• Data that is communicated to more than one consumer is deep copied automatically by the system, to avoid having the processing of one flow corrupt the status of another.