Age | Commit message (Collapse) | Author | Files | Lines |
|
Add IT test for read infrastructure + some additional unit tests
+ Make Read/Write Factory autoCloseable
Change-Id: I6eab8e6df2c2132af01cea0a9c4b9bece7dc9b74
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
+ Extract common registry builder base code
(Reader registry is not flat, so there is not full control over ordering as with writers
but it is sufficient)
+ Split CompositeReader into CompositeReader, SubtreeReader and GenericReader
+ No need to build composite structure in ReaderFactories (registry does that internally)
+ Keep only ReaderCustomizer + ListReaderCustomizer, no root reader (same for writers)
Change-Id: Ic4e5bc96ad47a6cbcada4efcc2209db5c16d2a6c
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
Now the registry is flat and allows for full control of writer execution order
Change-Id: I864e1d676588ffe59b596145e0829e81b1a1ed2f
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
Order of delete/write/update operations can have side-effects for
devices like VPP:
Adding more than one IPv4 address can in some circumstances
be silently ignored by VPP, therefore in case of updating IPs list,
addresses have to be deleted first.
Change-Id: I8712827e3c95232dac20a9ee35be1bc5b567bd54
Signed-off-by: Marek Gradzki <mgradzki@cisco.com>
|
|
+ Fix update subtree, child writer lookup
+ Change initializers operation to merge
Change-Id: I6ece7eb3d17d5a0b4a413189ddd383567d7e2270
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
Change-Id: I7b2b888fd7debb0aec3292a07fc35c0e6493d117
Signed-off-by: Marek Gradzki <mgradzki@cisco.com>
|
|
+ Fix 1 interface reads. Interface reader worked only
if GET interfaces-state was executed
+ Fix readSubtree for augmentations. Comoposite readers did
not check child readers for augmentations, only direct children.
Change-Id: I2bc433e3e5785453062ab262b9edabc72c333bf0
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
Change-Id: Ic75221fe61a0bb35753dd6f3b4deec1b8c020974
Signed-off-by: Marek Gradzki <mgradzki@cisco.com>
|
|
+ Remove dedicated reads to netconf-state in operational
datatree. Use regular reader from now on
Change-Id: I3d7245cd6cb5d9cf057aea600fddceb49cf18039
Signed-off-by: Marek Gradzki <mgradzki@cisco.com>
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
With broker, context data can be accessed in a transactional
manner, same as config data
+ Renamed data-api concepts to not include DataTree
+ Renamed context related concepts to better distinguish between them
+ Now passing full ReadContext to read customizers
+ Naming context is backed by context data broker
Change-Id: I0b2876dd74a31a9ced7d9b5145672868e12f8b82
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
+ Pass full WriteContext to the write customizers, making
them more flexible
Change-Id: I75c55aed02f9300eee20eabb4a3e84e294ed1e0f
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
Change-Id: Iebb52d59e2e342151e616548722366985d7ee9a5
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
There is a bug in ODL when checking auguments for equality:
https://git.opendaylight.org/gerrit/#/c/37719
Until the bug is fixed, updateCurrentAttributes will be invoked
even if there was no data change. It is customizer responsibility
to perform equality check if invoking update without actual data
change can cause side-effects in VPP.
Change-Id: Id72ec55ea7179330a4ebb1df49c29195908c5461
Signed-off-by: Marek Gradzki <mgradzki@cisco.com>
|
|
Change-Id: Ifc7d511b73c18e9ba0813c9321b475a825ec9af6
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
The order of execution was not correct
+ CompositeWriter ignored some changes
Change-Id: I53fd9fda4b7a0379e0fa8451fa894865f67ebace
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
Signed-off-by: Marek Gradzki <mgradzki@cisco.com>
|
|
Change-Id: Iffde1ee21b749a8c8efa96c243f7f8bb0811b4c7
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|
|
Change-Id: I281db366a112edc08203e8cb392a212708d4552a
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
|