Satisfy extension dependency by one of multiple extensions
Hello list,
I have a use case where an extension dependency can be satisfied by one
of five other extensions. Currently I'm unable to express that in the
extension control file, since the elements from 'requires' are currently
searched on exact name match. The attached patch changes this behaviour
for list elements that end with a *, into prefix matching, so that e.g.
table* matches tablefunc.
This allows me to specify in a controlfile
requires 'vocab*'
which is satisfied by having either one of the following extensions loaded:
vocab2005
vocab2006
vocab2008
vocab2009
vocab2010
thoughts?
regards,
Yeb Havinga
--
Yeb Havinga
http://www.mgrid.net/
Mastering Medical Data
Attachments:
extension_dependency_choice.patchtext/x-patch; name=extension_dependency_choice.patchDownload
diff --git a/src/backend/commands/extension.c b/src/backend/commands/extension.c
new file mode 100644
index 278bbcf..5bf7f8e
*** a/src/backend/commands/extension.c
--- b/src/backend/commands/extension.c
*************** static void ApplyExtensionUpdates(Oid ex
*** 114,145 ****
Oid
get_extension_oid(const char *extname, bool missing_ok)
{
! Oid result;
! Relation rel;
SysScanDesc scandesc;
HeapTuple tuple;
ScanKeyData entry[1];
rel = heap_open(ExtensionRelationId, AccessShareLock);
ScanKeyInit(&entry[0],
Anum_pg_extension_extname,
! BTEqualStrategyNumber, F_NAMEEQ,
CStringGetDatum(extname));
! scandesc = systable_beginscan(rel, ExtensionNameIndexId, true,
! SnapshotNow, 1, entry);
! tuple = systable_getnext(scandesc);
! /* We assume that there can be at most one matching tuple */
! if (HeapTupleIsValid(tuple))
! result = HeapTupleGetOid(tuple);
! else
! result = InvalidOid;
! systable_endscan(scandesc);
heap_close(rel, AccessShareLock);
if (!OidIsValid(result) && !missing_ok)
--- 114,152 ----
Oid
get_extension_oid(const char *extname, bool missing_ok)
{
! Oid result = InvalidOid;
! Relation rel, idx;
SysScanDesc scandesc;
HeapTuple tuple;
ScanKeyData entry[1];
+ int len = strlen(extname) - 1;
+ bool wildcard = (extname[len] == '*');
rel = heap_open(ExtensionRelationId, AccessShareLock);
+ idx = index_open(ExtensionNameIndexId, AccessShareLock);
ScanKeyInit(&entry[0],
Anum_pg_extension_extname,
! BTGreaterEqualStrategyNumber,
! F_NAMEGE,
CStringGetDatum(extname));
! scandesc = systable_beginscan_ordered(rel, idx,
! SnapshotNow, 1, entry);
! if (HeapTupleIsValid(tuple = systable_getnext_ordered(scandesc, ForwardScanDirection)))
! {
! Form_pg_extension form = (Form_pg_extension) GETSTRUCT(tuple);
! if (wildcard
! ? (strncmp(extname, NameStr(form->extname), len) == 0)
! : (strcmp(extname, NameStr(form->extname)) == 0))
! result = HeapTupleGetOid(tuple);
! }
! systable_endscan_ordered(scandesc);
+ index_close(idx, AccessShareLock);
heap_close(rel, AccessShareLock);
if (!OidIsValid(result) && !missing_ok)
On 23.09.2011 14:56, Yeb Havinga wrote:
I have a use case where an extension dependency can be satisfied by one
of five other extensions. Currently I'm unable to express that in the
extension control file, since the elements from 'requires' are currently
searched on exact name match. The attached patch changes this behaviour
for list elements that end with a *, into prefix matching, so that e.g.
table* matches tablefunc.
That's going to quickly extend into even more complex rules, like "foo
OR bar", "(foo OR bar) AND (foobar)" etc. IIRC the extension control
file format was modeled after some other package management system (.deb
?). You might want to look at the past discussions when the extension
control file format was decided.
We might want to have a system where an extension can declare that it
"provides" capabilites, and then have another extension "require" those
capabilities. That would be a neater solution to the case that there are
multiple extensions that all provide the same capability.
--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com
On 2011-09-23 14:19, Heikki Linnakangas wrote:
On 23.09.2011 14:56, Yeb Havinga wrote:
I have a use case where an extension dependency can be satisfied by one
of five other extensions. Currently I'm unable to express that in the
extension control file, since the elements from 'requires' are currently
searched on exact name match. The attached patch changes this behaviour
for list elements that end with a *, into prefix matching, so that e.g.
table* matches tablefunc.That's going to quickly extend into even more complex rules, like "foo
OR bar", "(foo OR bar) AND (foobar)" etc. IIRC the extension control
file format was modeled after some other package management system
(.deb ?). You might want to look at the past discussions when the
extension control file format was decided.
Ech.. 2364 hits on 'extension' in my mailbox. However I found a thread
'extension dependency checking' that also talks about version numbers
and the 'provides' capability you mention below.
We might want to have a system where an extension can declare that it
"provides" capabilites, and then have another extension "require"
those capabilities. That would be a neater solution to the case that
there are multiple extensions that all provide the same capability.
Yes that would be neater. I can't think of anything else however than to
add 'extprovides' to pg_extension, fill it with an explicit 'provides'
from the control file when present, or extname otherwise, and use that
column to check the 'requires' list on extension creation time.
--
Yeb Havinga
http://www.mgrid.net/
Mastering Medical Data
Yeb Havinga <yebhavinga@gmail.com> writes:
We might want to have a system where an extension can declare that it
"provides" capabilites, and then have another extension "require" those
capabilities. That would be a neater solution to the case that there are
multiple extensions that all provide the same capability.
+1
Yes that would be neater. I can't think of anything else however than to add
extprovides' to pg_extension, fill it with an explicit 'provides' from the
control file when present, or extname otherwise, and use that column to
check the 'requires' list on extension creation time.
That sounds like a good rough plan.
Then we need to think about maintenance down the road, some releases
from now we will need more features around the same topic. Debian
control file also has Conflicts and Replaces entries, so that using the
three of them you can handle a smooth upgrade even when the extension
changed its name or has been superseded by a new one which often has the
advantage of being maintained.
Regards,
--
Dimitri Fontaine
http://2ndQuadrant.fr PostgreSQL : Expertise, Formation et Support
All,
We might want to have a system where an extension can declare that
it
"provides" capabilites, and then have another extension "require"
those
capabilities. That would be a neater solution to the case that
there are
multiple extensions that all provide the same capability.+1
As a warning, this is the sort of thing which DEB and RPM have spent years implementing ... and still have problems with. Not that we shouldn't do it, but we should be prepared for the amount of troubleshooting involved, which will be considerable.
--Josh Berkus