File based

doesn't scale well

demlimited text e.g., CSV TSV
- easy to create and process
- portable
- lowest common denominator
structured text e.g., JSON, XML, YAML
- many tools for querying and transforming data
- portable also
Serialiased data Structures (usually binary)
- more compact
- easy to do
- single user only
- no automatic failure recovery
- no querying
- versioning issues
- no standards
- less portable

Databases

most dbmss have a native datbase api
- usually proprietry and limited to just that product
- often the only option for noSQL dbmss
also generic database apis
- work with multiple dbmss
- same code works with any supported dbms
for sql dbmss
- ODBC microsoft
- JDBC java
- DB-API python
- PDO php
- DBI perl

domain objects interact directly with the data store
- write to file or send sql statements
- not easy to change
domain objects interact with data store via a mediator
- either standalone class or implementation of a data access interface
- data access objects DAOs
- encapsulates all access to persistent data

general rule: one DAO per "logical unit" of data access
many DAOs are just for one class e.g., PatronDAO
some involve many classes
- things like header/lines objects are always managed together
- complex operations that join multiple tables or domain classes
different use cases (features) use different sets of DAOs e.g.,
- add, find, edit patron ⇒ PatronDAO
- lend items ⇒ LoanDAO, ItemDAO, PatronDAO
object construction and deconstruction coded into DAOs