Summary of team discussion on September 6, 2017
- Persist between sessions will be achieved by saving the data into a file, SQLite file format is one candidate.
- Firefly DataGroup object will remain but only be a value object. The operations on the data will be achieved through DB operations and/or stored procedures.
- Dealing with expressions:
- current expressions in plotting: evaluate them first, a parser and translator needs to be written to convert them to SQL form
- filter in table: Continue to support the simple column filter. complex ones will be in SQL form, with more attention on UI design
- One metaData table for each data table, with information:
- column name, type, format, ... Use IRSA DD table as reference
- special database if not H2 (see 'Keep options open' bullet)
- Keep options open:
- FF server directly access other DB, i.e. LSST user database in Science Platform.
- FF server uses DAX API for LARGE table data operations, instead getting the data into FF H2 DB.
- FF server access multiple database systems. Meta data for those special data tables will keep those information.
- Good practice
- cleanup the extra tables afterwards
- Future benefits:
- possible join of two tables
- Performance check:
- after the first round of implementation, performance comparison with the old system needs to be be done for up to ~ 1 million row tables.
- ?? billion row table is in a separate category
- Todo:
-
-
DM-11814Getting issue details...
STATUS
has been created for the first round of implementation
- one search processor update with new architecture, include ingesting returned results into H2
- data type mapping from current IPAC table to H2
- unique table ID generation, unique metaData table name/ID generation
- second round:
- other developers follow the guidance to work on more search processors
-
-
DM-11814Getting issue details...
STATUS
has been created for the first round of implementation