Opaque message parsing

A data processing system comprising at least one processor coupled to a memory having program code that is configured to perform, when executed by the at least one processor, steps of:

• evaluating program code for processing the message to identify a first set of data fields of the message that are referenced in the program code;
• identifying boundaries of the first set of data fields in a schema defining a format of the message;
• identifying a second set of data fields in the schema, the second set of data fields being related to the first set of data fields by reference, the second set of data fields further including the first set of data fields;
• and
• sequentially parsing the message using the boundaries of the first set of data fields, wherein sequentially parsing comprises skipping, according to the sequence, a subset of data fields of the first set data fields that precede in the sequence a first occurrence of a data field belonging to the second set of data fields, wherein the program code comprises a conditional expression including a plurality of branches, each of said branches referencing a different one of said data fields, said branch decision depending on a further data field downstream in said sequence relative to at least some of the data fields in said branches, the method further comprising:
• collecting run-time statistics from the parsing of a plurality of messages to determine the frequency of each branch being taken;
• selecting branches that are taken at a frequency above a defined threshold;
• identifying the data fields referenced by the selected branches;
• and
• skipping the parsing of data fields referenced by unselected branches that precede the data fields of the selected branches in said sequence.