Обсуждение: Class dependencies
Hi, Is it safe to assume all objects of a given class can be dropped/created, provided all objects of a list of other classes have already been dropped/created? I'm looking at http://developer.postgresql.org/pgdocs/postgres/catalogs.html For each class, a list of "References" are defined, i.e. other classes the given class depend on. For instance, is it correct to assume constraints always can be dropped, i.e. no other class (nor other constraints) can depend on them? -- Best regards, Joel Jacobson Glue Finance
On 2011-01-10, Joel Jacobson <joel@gluefinance.com> wrote: > Hi, > > Is it safe to assume all objects of a given class can be > dropped/created, provided all objects of a list of other classes have > already been dropped/created? > > I'm looking at http://developer.postgresql.org/pgdocs/postgres/catalogs.html > > For each class, a list of "References" are defined, i.e. other classes > the given class depend on. > For instance, is it correct to assume constraints always can be > dropped, i.e. no other class (nor other constraints) can depend on > them? As I unserstand it a references constraint requires a unique constraint on the referred-to expressiom. table a (b,c) references d(e,f) requires unique (e,f) on table d -- ⚂⚃ 100% natural
Jasen Betts <jasen@xnet.co.nz> writes: > On 2011-01-10, Joel Jacobson <joel@gluefinance.com> wrote: >> Is it safe to assume all objects of a given class can be >> dropped/created, provided all objects of a list of other classes have >> already been dropped/created? >> For instance, is it correct to assume constraints always can be >> dropped, i.e. no other class (nor other constraints) can depend on >> them? > As I unserstand it a references constraint requires a unique > constraint on the referred-to expressiom. Another problem for this type of scheme is circular dependencies. There are for example circular dependencies between a type and its I/O functions. pg_dump contains some heuristics for resolving the kinds of circular dependencies that are known to exist. regards, tom lane