[PATCH] [RFC] make_ref() of associations
# HG changeset patch # User Heidi Eckhart <heidieck@linux.vnet.ibm.com> # Date 1196686834 -3600 # Node ID bea3e027d42ff41fc452935427981739bab76573 # Parent a1582d092f517919470b9ce7ff034b89e4b2bade [RFC] make_ref() of associations While thinking about the implementation of make_ref() and about my proposal to base the asscociation's instance creation on a connect_by_classname, I came to the conclusion that opening a connection to libvirt for not using it, is overkill for this method. Once make_ref() gets called by the std_association logic, the provider can rely on that the prefix of the given reference was checked for the right hypervisor prefix. So I suggest to update all make_ref() functions to use get_typed_instance(). ... and refix some of my fixes :0. Besides that I suggest to rename some of the variables in make_ref() to make the relations clearer. Signed-off-by: Heidi Eckhart <heidieck@linux.vnet.ibm.com> diff -r a1582d092f51 -r bea3e027d42f src/Virt_ElementAllocatedFromPool.c --- a/src/Virt_ElementAllocatedFromPool.c Mon Dec 03 13:49:01 2007 +0100 +++ b/src/Virt_ElementAllocatedFromPool.c Mon Dec 03 14:00:34 2007 +0100 @@ -241,27 +241,27 @@ static CMPIStatus pool_to_vdev(const CMP return s; } -static CMPIInstance *make_ref(const CMPIObjectPath *ref, - const CMPIInstance *inst, +static CMPIInstance *make_ref(const CMPIObjectPath *source_op, + const CMPIInstance *target_inst, struct std_assoc_info *info, struct std_assoc *assoc) { - CMPIInstance *refinst = NULL; - - refinst = get_typed_instance(_BROKER, - CLASSNAME(ref), - "ElementAllocatedFromPool", - NAMESPACE(ref)); - - if (refinst != NULL) { - CMPIObjectPath *instop; - - instop = CMGetObjectPath(inst, NULL); - - set_reference(assoc, refinst, ref, instop); - } - - return refinst; + CMPIInstance *assoc_inst = NULL; + + assoc_inst = get_typed_instance(_BROKER, + CLASSNAME(source_op), + "ElementAllocatedFromPool", + NAMESPACE(source_op)); + + if (assoc_inst != NULL) { + CMPIObjectPath *target_op; + + target_op = CMGetObjectPath(target_inst, NULL); + + set_reference(assoc, assoc_inst, source_op, target_op); + } + + return assoc_inst; } char* antecedent[] = {
Heidi Eckhart wrote:
# HG changeset patch # User Heidi Eckhart <heidieck@linux.vnet.ibm.com> # Date 1196686834 -3600 # Node ID bea3e027d42ff41fc452935427981739bab76573 # Parent a1582d092f517919470b9ce7ff034b89e4b2bade [RFC] make_ref() of associations
While thinking about the implementation of make_ref() and about my proposal to base the asscociation's instance creation on a connect_by_classname, I came to the conclusion that opening a connection to libvirt for not using it, is overkill for this method. Once make_ref() gets called by the std_association logic, the provider can rely on that the prefix of the given reference was checked for the right hypervisor prefix. So I suggest to update all make_ref() functions to use get_typed_instance(). ... and refix some of my fixes :0. Besides that I suggest to rename some of the variables in make_ref() to make the relations clearer. Signed-off-by: Heidi Eckhart <heidieck@linux.vnet.ibm.com>
diff -r a1582d092f51 -r bea3e027d42f src/Virt_ElementAllocatedFromPool.c --- a/src/Virt_ElementAllocatedFromPool.c Mon Dec 03 13:49:01 2007 +0100 +++ b/src/Virt_ElementAllocatedFromPool.c Mon Dec 03 14:00:34 2007 +0100 @@ -241,27 +241,27 @@ static CMPIStatus pool_to_vdev(const CMP return s; }
-static CMPIInstance *make_ref(const CMPIObjectPath *ref, - const CMPIInstance *inst, +static CMPIInstance *make_ref(const CMPIObjectPath *source_op, + const CMPIInstance *target_inst, struct std_assoc_info *info, struct std_assoc *assoc) { - CMPIInstance *refinst = NULL; - - refinst = get_typed_instance(_BROKER, - CLASSNAME(ref), - "ElementAllocatedFromPool", - NAMESPACE(ref)); - - if (refinst != NULL) { - CMPIObjectPath *instop; - - instop = CMGetObjectPath(inst, NULL); - - set_reference(assoc, refinst, ref, instop); - } - - return refinst; + CMPIInstance *assoc_inst = NULL; + + assoc_inst = get_typed_instance(_BROKER, + CLASSNAME(source_op), + "ElementAllocatedFromPool", + NAMESPACE(source_op)); + + if (assoc_inst != NULL) { + CMPIObjectPath *target_op; + + target_op = CMGetObjectPath(target_inst, NULL); + + set_reference(assoc, assoc_inst, source_op, target_op); + } + + return assoc_inst; }
char* antecedent[] = {
_______________________________________________ Libvirt-cim mailing list Libvirt-cim@redhat.com https://www.redhat.com/mailman/listinfo/libvirt-cim
Couple of things here. First thing, I'm okay with renaming things to make it more clear, but I'm not really a fan of using "op" for ObjectPath. I realize it's a more logical abbreviation than "ref", but we use "ref" in so many places that I think it's pretty firmly established, and I personally always see "op" and think "operand" or "operation." I would recommend a compromise of "source_ref" for that one. Second thing, I think I'm missing something in the first paragraph of your commit message. I'm not entirely sure I understand the explanation that leads up to "So I suggest to update all make_ref() functions to use get_typed_instance()," and it also doesn't seem to really relate to this patch, which only appears to do some renaming in make_ref. It seems like a good point of discussion, and clarification would be good, but maybe we should move it to a separate thread and limit this patch discussion to the variable naming. -- -Jay
JG> I realize it's a more logical abbreviation than "ref", but we use JG> "ref" in so many places that I think it's pretty firmly JG> established For the parameter, I think source_ref would be fine. I don't think changing it to not include "ref" makes it more clear, since the convention (as it seems to me) is that the object path you pass in as the "subject" of an operation is called the "reference". For the "refinst" local variable, I think leaving "ref" in the name also makes sense. The goal of the function is to "make a reference", so "refinst" seems to be a reasonable name :) JG> I personally always see "op" and think "operand" or "operation." I JG> would recommend a compromise of "source_ref" for that one. Me too! I'd be okay with "source_ref", but I question how much more clarity it will bring given how much patch churn it will produce. However, if others want this change, I'm okay with it. -- Dan Smith IBM Linux Technology Center Open Hypervisor Team email: danms@us.ibm.com
Dan Smith wrote:
JG> I realize it's a more logical abbreviation than "ref", but we use JG> "ref" in so many places that I think it's pretty firmly JG> established
For the parameter, I think source_ref would be fine. I don't think changing it to not include "ref" makes it more clear, since the convention (as it seems to me) is that the object path you pass in as the "subject" of an operation is called the "reference".
For the "refinst" local variable, I think leaving "ref" in the name also makes sense. The goal of the function is to "make a reference", so "refinst" seems to be a reasonable name :)
JG> I personally always see "op" and think "operand" or "operation." I JG> would recommend a compromise of "source_ref" for that one.
Me too! I'd be okay with "source_ref", but I question how much more clarity it will bring given how much patch churn it will produce. However, if others want this change, I'm okay with it. Thanks for the comments. I was resending the patch as "variable renaming in make_ref() of associations".
Yes, its very good to question this patch. And the more I think about, the more I find, that there is the possibility to generalize the content of make_ref() - because at the end most of the association's make_ref() will look the same, besides the different association class names. I will send out the patch for this idea. -- Regards Heidi Eckhart Software Engineer Linux Technology Center - Open Hypervisor heidieck@linux.vnet.ibm.com ************************************************** IBM Deutschland Entwicklung GmbH Vorsitzender des Aufsichtsrats: Martin Jetter Geschaeftsfuehrung: Herbert Kircher Sitz der Gesellschaft: Boeblingen Registergericht: Amtsgericht Stuttgart, HRB 243294
Jay Gagnon wrote:
Heidi Eckhart wrote:
# HG changeset patch # User Heidi Eckhart <heidieck@linux.vnet.ibm.com> # Date 1196686834 -3600 # Node ID bea3e027d42ff41fc452935427981739bab76573 # Parent a1582d092f517919470b9ce7ff034b89e4b2bade [RFC] make_ref() of associations
While thinking about the implementation of make_ref() and about my proposal to base the asscociation's instance creation on a connect_by_classname, I came to the conclusion that opening a connection to libvirt for not using it, is overkill for this method. Once make_ref() gets called by the std_association logic, the provider can rely on that the prefix of the given reference was checked for the right hypervisor prefix. So I suggest to update all make_ref() functions to use get_typed_instance(). ... and refix some of my fixes :0. Besides that I suggest to rename some of the variables in make_ref() to make the relations clearer. Signed-off-by: Heidi Eckhart <heidieck@linux.vnet.ibm.com>
<...> Second thing, I think I'm missing something in the first paragraph of your commit message. I'm not entirely sure I understand the explanation that leads up to "So I suggest to update all make_ref() functions to use get_typed_instance()," and it also doesn't seem to really relate to this patch, which only appears to do some renaming in make_ref. It seems like a good point of discussion, and clarification would be good, but maybe we should move it to a separate thread and limit this patch discussion to the variable naming.
Sorry for causing this confusion. The renaming was one thing, but as the patch showed how the code for each make_ref() would look like, I wanted to also bring up this idea. But not very well described. I will send out a patch "#2 - variable renaming in make_ref() of associations", which shows what I meant. -- Regards Heidi Eckhart Software Engineer Linux Technology Center - Open Hypervisor heidieck@linux.vnet.ibm.com ************************************************** IBM Deutschland Entwicklung GmbH Vorsitzender des Aufsichtsrats: Martin Jetter Geschaeftsfuehrung: Herbert Kircher Sitz der Gesellschaft: Boeblingen Registergericht: Amtsgericht Stuttgart, HRB 243294
participants (3)
-
Dan Smith -
Heidi Eckhart -
Jay Gagnon