sql_build()
creates a select_query
S3 object, that is rendered
to a SQL string by sql_render()
. The output from sql_build()
is
designed to be easy to test, as it's database agnostic, and has
a hierarchical structure. Outside of testing, however, you should
always call sql_render()
.
join_query(x, y, vars, type = "inner", by = NULL, suffix = c(".x",
".y"))select_query(from, select = sql("*"), where = character(),
group_by = character(), having = character(),
order_by = character(), limit = NULL, distinct = FALSE)
semi_join_query(x, y, anti = FALSE, by = NULL)
set_op_query(x, y, type = type)
sql_build(op, con = NULL, ...)
sql_render(query, con = NULL, ..., bare_identifier_ok = FALSE)
sql_optimise(x, con = NULL, ...)
A sequence of lazy operations
A database connection. The default NULL
uses a set of
rules that should be very similar to ANSI 92, and allows for testing
without an active database connection.
Other arguments passed on to the methods. Not currently used.
Is it ok to return a bare table identifier.
Set to TRUE
when generating queries to be nested within other
queries where a bare table name is ok.
sql_build()
is generic over the lazy operations, lazy_ops,
and generates an S3 object that represents the query. sql_render()
takes a query object and then calls a function that is generic
over the database. For example, sql_build.op_mutate()
generates
a select_query
, and sql_render.select_query()
calls
sql_select()
, which has different methods for different databases.
The default methods should generate ANSI 92 SQL where possible, so you
backends only need to override the methods if the backend is not ANSI
compliant.