git.opendaylight Code Review - yangtools.git/log

Add Abstract{LeafSet,Map}ModificationStrategy

This allows us to remove some code duplication in constructor,
but notably share the implementation of getChild().

Furthermore it allows us to make the rooted-at-MapEntry hack work
consistently for ordered map nodes too.

JIRA: YANGTOOLS-946
Change-Id: Id91529c31341afd46a693f24eb1ac4ea0dc4800d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 73740a4c50094c722152883bd333e82961c4dd4c)

Convert MinMaxElementsValidation to ModificationApplyOperation

This removes the fragile piggy-backing onto SchemaAwareOperation,
as done as part of YANGTOOLS-909 and modifying it to take advantage
of DelegatingModificationApplyOperation. This allows us to hide
SchemaAwareApplyOperation's checkWriteApplicable().

Change-Id: I804ddb5e4c8b7562bde11994be8b1a5615228d7c
JIRA: YANGTOOLS-944
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Introduce DelegatingModificationApplyOperation

As it turns out implementations outside of SchemaAwareApplyOperation
can share implementations of most methods through delegation. Create
two new abstract classes, DelegatingModificationApplyOperation and
FullyDelegatedModificationApplyOperation which host partial and full
set of methods implemented as delegates.

FullyDelegatedModificationApplyOperation is used by AlwaysFailOperation
and RootModificationApplyOperation, as they the always-fail behaviour
is easily implemented through throwing ISE in delegate().

This eliminates 1-2 implementations of each ModificationApplyOperation
method.

Change-Id: I578769418baafb285593fefe8868e97e1d171157
JIRA: YANGTOOLS-944
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 37fe9edc082b151bc100537cfa567ba5998d453b)

Cleanup AbstractValueNodeModificationStrategy

The implementation did not log offending values and reused
verifyStructure() to perform validation, making analysis difficult.

Refactor the value check so that it is in a private method invoked
from the three codepaths which need to perform it, making code flow
clearer.

Change-Id: I315899982c0f304e0459d5ad33a19e56af0d2355
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 59227494c2ba2ff1543874b32e69cd5f44d9d51c)

Fix automatic lifecycle delete stacking

Nodes with automatic lifecycle have the strange property that their
delete operations may actually stack with other operations.

Unfortunately delete operation itself has no semantics of having
underlying modifications, as it has no baseline to which to apply
them.

In order to fix this, we need to turn deletes into empty writes
for these constructs and once we are performing disappearing, undo
this trick. This effectively means that both empty writes and deletes
result in a delete operation, if they ended up modifying the tree,
or into a no-op if they did not.

JIRA: YANGTOOLS-938
Change-Id: I68b9f304e76d6b1bdb707f411a87cd6085ed8bc7
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Optimize StructuralContainerModificationStrategy.fakeMeta()

The container node we use for fakeMeta can easily be shared, which
improves our performance profile.

Change-Id: I5a7afc1ea9a9f66d7556e8e6b2b9ef79df9e5e57
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Update junit link

javadoc.io breaks with JDK11, update the link.

Change-Id: Ia3a1bb2905a0e55797514537f220332754515c1d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 9ecf2b5f7f3049ad7b7e2efca74d058c3d41899e)

Fix artifacts typo

yangtools-docs is at the same version as yangtools, i.e. 2.0.17,
not 2.0.9.

Change-Id: I3d2473cee9d8b884931fc8a1960d992d2cd0d690
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 8dd2957dc9919df381b8c413600bdbb58823fcd3)

Fix yangtools-docs

Current docs fail to correctly deploy, fix that up. Also group
packages to make them more readable.

Change-Id: Iebf8c785ca0fd7968724759a2b593ca559620057
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit d15579af537c873f1b0e4f2954af6b5db962ba76)

Rework aggregate javadocs

Since javadoc:aggregate-jar is forcing a full re-generate and compile
of the project, let's take a different approach.

This patch overloads the purpose of the docs directory so that it
participates on maven build. It produces a pom file with attached
sources and javadocs. Later we can add more stuff to it, too.

Change-Id: I866e8405e73f297c552194a7caf10c1fd8b26731
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 14372ca8ee0e2da543549143cba4eedb585530b6)

Allow {Json,Xml}ParserStream to handle actions

Generalize the support for RPCs to include actions, so that we can
use both. Also fix null handling to comply with javadoc.

Change-Id: Id84052c1d1f20aa9d0371015f01b0c8845152de2
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 80ae3a4d7f12024341ef4900b4d672c17abd0262)

Allow JSON/XML writers to be instantiated with root node

Rather than requiring a SchemaContext/SchemaPath combination, allow
writers to be instantiated by specifying a DataNodeContainer.

Change-Id: I9eb470587cad2af9b3390d134dcc03abc95e8645
JIRA: YANGTOOLS-935
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit aa37f2a86bfb4d04a99f78b1b0d892b3c0b76caf)

Refactor implementations to hide XMLStreamWriter

In order to have control over when we output an element, we need
to make sure we know how the writer is being used.

Refactor XMLStreamNormalizedNodeStreamWriter interaction with its
subclasses so we know when we are leaking the writer -- which
happens only in writeValue().

JIRA: YANGTOOLS-927
Change-Id: I7e7bd62016e994a31fe6d69995e9d088746359da
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 127cfbef06dec06721efc4fafe2e9ab551c644e6)

Bump versions for 2.0.17-SNAPSHOT

This starts the next release development.

Change-Id: I0f19deaf1c7f1d12107f9880b834d63d1de9d8b7
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump to odlparent 3.1.6

Change-Id: Id89e75d896a1f9863afec520f8dce47a6fe4eef6
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Do not write empty string in EmptyXmlCodec

Emitting an empty string does nothing but churn writer state,
so speed up EmptyXmlCodec by not touching the writer.

Change-Id: Iab7db37089b84ec1a3c2e3c9381265025e21fdca
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix default namespace emission

When we are emitting opening a new element, we need to consult
XMLStreamWriter before we emit the element, otherwise its state
already reflects the newly-started element.

In non-repairing mode the JDK writer will accept our invocation
and update the namespace mapping, which will lead us to erroneously
assume we do not need to emit the declaration.

Since it is non-repairing, it will not emit the declaration itself,
leading to a blank document -- something for which we are actually
testing in one of our tests.

In order to improve the confidence in results, also update tests
which used only repairing factories to use both repairing and
non-repairing ones.

JIRA: YANGTOOLS-801
Change-Id: I6975cf21121f7a8f92056b5b49b31d53b685dc0f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 706aaed3e21eb3c5d080ee974f27b9348c3d2e92)

Improve JAXP implementation compatibility

Testing with woodstox-core shows a couple of UT failures that
are really assumptions on the implementation. Address them so we
do not get failures when implementation changes (too much).

While we are at it, improve startsWith/contains string checks
by using hamcrest.

Change-Id: I4af0c9cc370e5699255725f049ee7bb9b5c171ac
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit fea3a6991aa892b676ab550393e944c3e6215ef3)

Attach aggragate javadocs to aggragator

This allows us to point users to our aggregated javadocs, hence
they do not fish for them.

Change-Id: If191cdfd45c2732a96d4a62deb183f6add155a65
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 5319ee7804ebecd7bf022e2a1c1df117945834fa)

Shorten StatementMap$Regular.get(int)

We can use a trigraph instead of a if/return combo, saving an
instruction.

Change-Id: If1482da5a4cf83f87b4d995333d757d2fa2d7e12
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix StatementMap$RegularAsCollection's size

We are not initializing the size field, hence the class does not
operate correctly. Fix that and correct isEmpty() method.

Change-Id: I1543e617f14f0a88c6b59afe93364cace151ee5c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit d94cd2529378a5544030c75bdb23b05d98092e05)

Cleanup RFC6020_YIN_NAMESPACE usage

In export write we really want to use RFC6020_YIN_NAMESPACE_STRING,
not RFC6020_YIN_NAMESPACE, making references proper constants.

Change-Id: I3e9c7c745fbd2dd97fcb35e47cdb0920a81cc596
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit a298f385bbb5a1ca9aeca03516bfac81c22bb1fd)

Fix noRev map efficiency

QName.getRevision() always returns non-null, which means we are
always taking the slow version of this branch, as correctly flagged
by Eclipse.

Furthermore we can use QName.withoutRevision() to side-step QName's
localName validation, as the local name is known to have already
been checked.

Change-Id: I7be3b8053ab3dbddfdf9cf12f8754a307127a8b3
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit e4aa6ef84602dfe47720fa6ebba8fff53bb6886f)

Speed up StatementContextVisitor.getValidStatementDefinition()

Unprefixed statements are most common and we end up creating
a QNameModule each time we encounter them, while we have a common
constant which we can use.

Use the constant, allowing us to allocate fewer objects while
also allowing making interning work.

Change-Id: I9e8847f8bd39fb9bcf6f8d0e8b86d196aeb5e2cd
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to .16-SNAPSHOT

This starts the 2.0.16 development iteration.

Change-Id: I36298775de0fa1e228f6386df17bc7ff3fe17b4e
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump to odlparent 3.1.5

Change-Id: I9c0a3b8c3c4b2c3d03bb8e3df623d87cd70d724b
Signed-off-by: Stephen Kitt <skitt@redhat.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add karaf distribution

This adds a karaf distribution for the purposes of validating
upstream (odlparent) packaging.

Change-Id: I0c01ebad2cad2e34407eae139cb6b2a2dc119d51
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit a0c2c8dee45b8fc2528d7d0e45461f215ca42b49)

Bump to odlparent-3.1.5-SNAPSHOT

This bumps to current development version for validation purposes.

Change-Id: I39cdf113bae58184e7601dc160dc97554c1ca280
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Intern empty leaves

This patch teaches LeafInterner to intern leaves holding Empty,
as they have only one possible value.

Change-Id: I5afa161ec49b02c33a3e7be28bcda135b6d796cb
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 2.0.16-SNAPSHOT

This starts the next development cycle.

Change-Id: Ibdcd4ded6a069a64899d52357efe64b4c175da4d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Preserve attributes when parsing AnyXml from XML

So far, anyxml nodes have been parsed in following pipeline:
String -> DOM -> Stax -> String -> DOM
Now the parsing is simplified to:
String -> DOM -> Stax -> DOM

Remove the unnecessary "back to string" step. This step is not just
unnecessary but also stripes down attributes and namespaces from AnyXml
node value.

JIRA: YANGTOOLS-923
Change-Id: If13ce9c7397c37b96a532090e67683a635ae14ef
Signed-off-by: Jakub Morvay <jmorvay@frinx.io>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 335359d23557e2b54b9df539eeaed8b97668a59e)

Allocate new keyes map lazily

Do not allocate the map for new mappings until it is actually needed.
This introduces some additional complexity via null checks, but that
is not something major.

Also refactor large equals() method into smaller chunks, so that it
can be inlined more easily.

JIRA: YANGTOOLS-919
Change-Id: Idb4b7a4cb7cde09af2eab0bcebfb71083e64ff2f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 62876a059465a1c780eb88e9ef5f285f4a052ec2)

Bump versions to 2.0.14-SNAPSHOT

This starts the next development iteration.

Change-Id: I14fd1f6e73edf822a5f250b8629aa48d6cfa1458
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Separate out ConstantArrayCollection iterator

This splits it out into its own static class.

Change-Id: I94a86c65eefdba79db2f5ec3a01b13dbad0b41b0
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add predicate-free NodeIdentifierWithPredicates constructor

This saves some time and figuring out to users, as they can just
not specify any predicates and get a free list. This side-steps
some of the code done in the Map-taking constructor.

Change-Id: If06c50aad3c2918cc6ec150b4185837557ac3f40
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 6112dc1bfb6245584005509d5805afde661d9047)

Refactor ImmutableMapEntryNodeSchemaAwareBuilder

This builder's logic is not entirely efficient, as it can constructs
unnecessary NodeIdentifierWithPredicates purely for checking purposes.

Furthermore we can side-step child validation when we are switching
by available children, as we have the entryset available.

JIRA: YANGTOOLS-917
Change-Id: I57d0a156ced3abce7e61403c2b390c074850d612
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit c43197e66be08a0ffaa00bbe8405fa4ba23b8656)

Factor out {SharedSingleton,ImmutableOffset}MapTemplate

Exposing the single-entry map case has the advantage of providing
instantiateWithValue(Object) method, side-stepping the need for
a temporary array. Split it out of ImmutableMapTemplate, so that
users can provide customized handling as needed.

The same holds true if the keyset is known to contain more than
one key, in which case using ImmutableOffsetMapTemplate can be
faster.

JIRA: YANGTOOLS-917
Change-Id: If08f492ff59bc89f3826a4122cfbcf34657c2210
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit bd16b6161802322cb33c39642ddeb8a59b3e5d48)

Improve ListEntryNodeDataWithSchema

ImmutableMapTemplate allows us to more efficiently instantiate
NodeIdentifierPredicates, as it takes care of ordering of keys
while maintaining an efficient internal representations.

It furthermore provides a keySet() method, which is useful for
quickly checking if a leaf is part of key definition -- unlike
ListSchemaNode.getKeyDefinition(), which is a List and therefore
must be linearly searched.

It also allows us to skip HashMap allocation for unkeyed lists,
clearly showing that the class should be split into two.

JIRA: YANGTOOLS-917
Change-Id: If2d7bbbc5c4fe1ca96097f374f4313aabb411908
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit e31b3dc700002b19d92ecc68463ab2711d8af99c)

Add ImmutableMapTemplate

We have couple some hot paths where we know the keySet of an immutable
map we want to instantiate. For these, using
ImmutableOffsetMap.*copyOf() is wasteful, as it performs unnecessary
cache looks and map lookups.

ImmutableMapTemplate can be used to amortize these preparatory steps,
so that an ImmutableOffsetMap (or SharedSingletonMap) can efficiently
be produced by either transforming values of some other map, or by
supplying values in the correct order.

Since we are in the area, improve documentation and @NonNull
annotations of both ImmutableOffsetMap and SharedSingletonMap.

JIRA: YANGTOOLS-917
Change-Id: I8e7808b9bfe751885b842715f7e0346c0e7485b2
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 048042bd88cc9b1f481c018363c331bca83e4e1f)

Speed up UnorderedMapMixinContextNode

NodeIdentifierWithPredicates recognizes ImmutableMap, make sure
we remain on the faster path of it.

Change-Id: I95d745d6e7ba54dcdb422fac19256112a21f413f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 6116391d4f24cac829f8cead8ee4426c5a0ad2bf)

Retrofit SchemaContextProvider

Since we already have a getSchemaContext() method, we can
get it documented and widely available for free.

Change-Id: Ib5aaa3d1d503a000a997beaee3d31002755e5998
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 5227b0accfa016d76d32ddad414cce1c62585664)

Make InMemoryDataTreeSnapshot implement SchemaContextProvider

Note DataTreeSnapshot's duplication of getSchemaContext(), which
is indicative of a coupling to yang.model.api. This is fine for now,
as this coupling already exists through xpath interfaces, but it
needs to be revisited and cleaned by either removing the method,
or making it a non-default (and inherited from SchemaContextProvider).

Change-Id: If335f178e9e086b8b5185044a06d4296f81b2a74
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit ccdb5a655b07693eedbe96d4d5215ff52cbfd671)

Refactor XMLStreamWriterUtils

This refactors XMLStreamWriterUtils, so that it can be a proper
SchemaContextProvider. This allows us to propagate this property
through SchemaAwareXMLStreamNormalizedNodeStreamWriter the same
way.

Change-Id: Ibe0832e76d3e66ca27f8d345f7fd6f890b92f35e
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 9c73f586bfae63818e0ded04dd8282d4c7c8e468)

Reuse writer field from parent

Writer is a duplicate field, which is already available as
getParent(). Remove the field and use the getter.

Change-Id: I1d5dd866512479268e32affb0ded301fadb5cb11
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Document Identifiable interface contract

Somehow this core interface went undocumented, fix that and also
clarify it's companion Identifier contract.

Change-Id: I009ed90e548770b46d636e5e41c07877642682d5
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 89ddc0a57c389ff3feb403cbd0bd5709f01a8e14)

Cleanup Immutable documentation

This is a simple formatting update, with small rewording to use
better language.

Change-Id: I51848c3a956c7251bad98418cac40ea0ffa943c2
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit c636aeaa87c46fdea600299f193f2cbf72c3f6ca)

Add OptionalBoolean.ofOptional()

This adds a simple utility method to create symmetry with
fromOptional(). Also cache TRUE/FALSE optionals, so accesses
get a bit faster.

Change-Id: Ife95d26f9b58e6859991d544eda52d0094302c30
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 520eb808fe7c2fba015be566f25f7ecf689cdf9f)

Properly propagate BuildContext

YangToSourcesProcessor was not properly propagating BuildContext,
fix that.

JIRA: YANGTOOLS-915
Change-Id: Id4f28fe640bbd4d4ed065b9aa62f0becb2db43ec
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 4bdd0b2f66433d7586a1eb32636bdab8a1d4361c)

Tolerate empty instance identifier in ImmutableNodes.fromInstanceId()

If the user specifies an empty InstanceIdentifier, we fail with an
undocumented NoSuchElementException. Make sure we properly handle
this case by returning a container which matches SchemaContext.NAME.

JIRA: YANGTOOLS-914
Change-Id: I0b503d3bed99d73d88e037d35979c503215f3b52
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Limit the TCCL when instantiating UntrustedXML

JAXP newInstance() methods can end up going through ServiceLoader
with TCCL as its ClassLoader, which means our constants may end up
pointing to a ClassLoader beyond our normal visibility.

This means we can end up picking up a different factory depending
on where we are called from and also can mean the TCCL cannot be
garbage-collected.

Force the TCCL to be the ClassLoader holding UntrustedXML, thus
establishing tight coupling between JAXP implementation discovery
mechanism and UntrustedXML.

JIRA: CONTROLLER-1867
Change-Id: Icacb0e193c7360c73bd07bc5c815bf803ca9ac44
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove unneeded jaxb-related exclusions

odlparent-3.1.4 contains proper feature exclusion and also
proper dependecyManagement declaration of jaxb-api, remove them
from here.

Change-Id: I6c04bb424da7bb413169fbf2eb54f7b387246a61
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add CopyHistory.toString()

This method makes for easier debugging, add it.

Change-Id: Ib25dc0414dce2ddacecaedafb14b2e0d42e3d50d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 2.0.13-SNAPSHOT

This starts the next development iteration.

Change-Id: I415dd6cb7730e2cd18bab96d69e33eb1bc9e4797
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump yang-xpath versions to 0.1.12-SNAPSHOT

This bump was missed previously, fix that up.

Change-Id: Ia5aa358097b989e80cdfdf4b47f8b0b2ce29cc2d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump to odlparent 3.1.4

This upgrades to the released version.

Change-Id: I8514dde60713ab8480ebb00bb74eac1f838f049a
Signed-off-by: Stephen Kitt <skitt@redhat.com>

Bump to odlparent 3.1.4-SNAPSHOT

This bumps our use of odlparent to 3.1.4-SNAPSHOT in anticipation
of a full release.

Change-Id: I8d03ed7321520e9580beb400534d20215d34edd5
Signed-off-by: Stephen Kitt <skitt@redhat.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Attach error listeners to lexers

We currently disconnect parser from standard output, but lexer
may incur errors as well, hence we should do the same dance to
lexers, too.

Change-Id: Ib901cb1fbdba05c721aeba5d60a5e8bda5f48087
JIRA: YANGTOOLS-905
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit b88d8bcfa64908ccd2b64462dc6b66bb3c69a47c)

Improve IfFeaturePredicateVisitor

We need to detect parser errors and not print them to stderr,
create an abstract class with the validation functionality
and use it when parsing an if-feature argument.

Change-Id: I63ad241bcfeeda7cb05f1472f593167d5ce08404
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 41cd74eff8a48f5437e76990319c97fd7bdd5df2)

Exclude javax.xml.stream dependency

This dependency is satisfied by the JDK, hence we do not need to
pull it in.

Change-Id: I440114204f064239e3e25215e53fe67981ddf6d7
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit a2c8f1405a44b41ccb17ec801beeaa78420336c5)

Update .gitreview to track the v2.0.x branch

Change-Id: Ibcc3381eb2349a91b088de186e8f497280e08a53
Signed-off-by: Stephen Kitt <skitt@redhat.com>

jaxb-api should be scope=compile

Keeping scope=provided breaks downstream tests with JDK9+,
hence we need to rework the dependencies to make sure we always
pull jaxb-api, but we also need to exclude it from odl-yangtools-data
feature, because it is provided by Karaf.

Change-Id: Id09281c89a74c3d6c780179dfa21ced545318ba1
JIRA: YANGTOOLS-906
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit eeb3076563314a67ccbe2d083dede2051d12e060)

Cleanup ScannedDependencyTest

Rather than messing with test resources set them up properly
in the project so that we can rely on them being present.

Also make sure we clean up our resources correctly.

Change-Id: Idb1b8a8d87293a255582f7120cdb0e47f7278080
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add NoOp{Listener,Object}Registration

As it turns out, there are use cases for registration which do
nothing in their close() method. Add utility classes to support
those use cases. Also update javadocs a bit.

Change-Id: Ife196f26229cf54173e9b9046ece22bd0138c2ea
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Add jaxb-api dependency

For JDK11 we need an explicit scope=provided dependency on
jaxb-api.

JIRA: YANGTOOLS-906
Change-Id: Ifee829a5cd04561f6124ff03b772db2f5d12f04e
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit aa676c9045ce081687f814607e977a0609c77bc5)

Do not clutter logs with a thrown exception

SourceException is an unchecked exception which we handle at a higher
layer. The logging statement does not add anything useful, so let's
remove it.

Change-Id: I8a4654b44469a2193996d1d90e982d01a1a62bf3
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix TypedefEffectiveStatementImpl double-checked load

We really need a volatile read here to make things alright, make
the field volatile.

Change-Id: Id04917a04f9f1fdba6ded2c773571d90106e877f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove unneeded interface declaration

DataNodeContainer is already specified by superclass, no need to
repeat that declaration.

Change-Id: I9a409662d87b02ec9b4abcba88c90bccf684829a
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Remove unneeded check

brk is known to be different from -1, hence we can use it directly
rather than using a temporary variable.

Change-Id: If089baa50b8e7b15d8ab03a8e09f7d6d6d0ea7e6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Bump versions to 2.0.12 release cycle

This bumps versions to next development release.

Change-Id: I427f07c850593195e9479a957984d5a71fef1864
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Fix logging template

Logger template is missing a placeholder, add it.

Change-Id: Ib1068bc4e122affc69e408da87378230607b7408
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit cac1831a621574716e403afe7997360308758fa6)

Consider AugmentationNodes when direct descendant is not found

With previous refactors it is quite easy to arrive at a simple
solution to take augmentations into account. This also

This includes a unit test from upstream based on afi-safi-name
with bgp-openconfig-extensions.

JIRA: YANGTOOLS-892
Change-Id: I237ed2c6cbed458a2109a48b468a30650f103a48
Signed-off-by: Marek Gradzki <mgradzki@cisco.com>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit e4777b6feff804cdcf633c28d4b654b0e885dbd1)

Eliminate LeafRefValidation.forEachChoice()

As it turns out the two call sites can share more code. Introduce
processChildNode(), inlining forEachChoice() and its surrounding
lookup logic.

JIRA: YANGTOOLS-892
Change-Id: I2091abcd579ce1d48a628c3bdea9b4514b050e29
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Split up LeafRefValidation.validateNodeData()

Turn this method into a dispatch and move logic into dedicated
methods. This makes it easier to follow the code flow and eliminate
duplicates.

JIRA: YANGTOOLS-892
Change-Id: I840fa0d2e8c7ea86efa05c1afea825e85d890837
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 3d22aba92422b08413d95b372ac11a2b26296a01)

Minor code cleanups

- static private method
- reduce use of ChoiceSchemaNode.getCaseNodeByName()
- QNameWithPredicateBuilder(QName)

Change-Id: I9cf97ba4c6ef667e39212f511e17bb18b4002bc3
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit ef795628f150556ddc8c93b865d1c3c0b9b0109c)

Use Iterables.filter() instead of FluentIterable

This removes allocation of one instance of FluentIterable, reducing
the amount of garbage we produce.

Change-Id: I93953f1263426f7fff20e579e56f82506fb423d4
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit ca61467e9c4e802584a1475da15c01cc5ea8e83a)

Rename LeafRefValidatation to LeafRefValidation

This is a naming typo which went unnoticed for quite some time,
rename the implementation class and instantiate a proxy for
compatibility.

JIRA: YANGTOOLS-892
Change-Id: I3ec77d7df41942502394047e9e13007c6693d3af
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 2d5b39317c3c5db3fa9c4c258175340d3e347c38)

Rework choice iteration in LeafRefValidatation

Rather than extracting ChoiceNodes into a temporary list, use
a consumer which will be invoked for each of them.

JIRA: YANGTOOLS-892
Change-Id: I0956bed6e446c1ee4bf44c88fcc361abe8a508e5
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 4a29fdd9dadf755ff6b3b6e34e9ff43c22a6b765)

Cleanup map iteration in LeafRefValidatation

We have two codepaths for non-filtered and filtered map entry case,
which end up iterating over map entries. Rewrite the code in terms
of an optionally-filtered stream, reducing code duplication.

JIRA: YANGTOOLS-892
Change-Id: I3ca37edb66373cc5ab4632210980cd4a467fd198
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Rework path LeafRefValidatation path tracking

Rather than using a plain iterable sourced from LeafRefPath.getPathFromRoot(),
maintain a Deque sourced via getPathTowardsRoot(). This eliminates the need
for Iterables.skip() as well as not forcing getPathFromRoot() being initialized.

Change-Id: I23c054a63b2c0ceed24fcf48d2636508ee13d568
JIRA: YANGTOOLS-892
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Pass down predicates instead of the entire predicate node

Since we are not using the rest of the data, it is much simpler
to just pass down predicates. It also reduces confusion as to
what state we are propagating.

Change-Id: If790fcc2a8d023cb863d09162b25d644b0dd4638
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit e8ecfb6f906dfd7e7bff791abf0b601eb71766a8)

Do not validate non-existent roots

If we have no data there is no point to validate anything, and
having the root value as an invariant makes code a lot simpler.

JIRA: YANGTOOLS-892
Change-Id: Ifd68b09c03d90c8881381809047e59f3fe92682b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit d2e80009970418f4755854f0918c5d80afb21a34)

Add LeafRefValidatation.computeValues()

This separates out the heavy-weight value computation, simplifying
callers a bit.

JIRA: YANGTOOLS-892
Change-Id: I218659f830aec822f80decd3622c9890df832379
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit dd1f03b367b7fcec4d3a0a10bc81d4126d7c88f4)

Eliminate leafRefTargetNodeDataLog()

Now that we captured invariants, we can proceed to turn
leafRefTargetNodeDataLog() into a simpler lambda form.

JIRA: YANGTOOLS-892
Change-Id: I4644d0f3bc31f9f9cb3643e8443159522f14ccaa
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 86e7fb7d83ea73633b4e50bfbe95ded122b69a85)

Untangle result processing logic.

leafRefTargetNodeDataLog() is needlessly complex, where it is called
with 3 different sets of arguments from validateLeafRefTargetNodeData()
only.

Inline two of those callers, reducing nesting significantly and
making the code flow more obvious.

Change-Id: I68205d58677e2392989a5fc064a986f5c1716666
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit a538bafb5db59101206a5d1fd2a1d31e79c08de0)

Move LeafRefPath creation from fast path

LeafRefValidation is the fast path and LeafRefPath we are creating
is in an invariant. Cache it in LeafRefContext as needed.

JIRA: YANGTOOLS-892
Change-Id: I40ef136630b7cdacca84b2f08a9e0491daa03c14
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 4b7393b11151722f44c19236684c661cf5f8edd0)

Cleanup leafref path parsing

We have a String on input and LeafRefPathParserImpl uses ANTLR's
CharStreams -- hence it makes no sense to bounce input through
an InputStream.

Remove this atrocity, eliminating an IOException/IllegalStateException
error path in process.

JIRA: YANGTOOLS-892
Change-Id: I78eadf30da6a1e64621ca4c14dfa13e84524b591
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 23b7ff7e2e39e0d520e249ae5b256746acef5abd)

Cleanup LeafRefContext(Tree)Builder

These are package-private classes, make them final, eliminate
unneeded methods and improve code formatting.

JIRA: YANGTOOLS-892
Change-Id: I8a4925a665b1bcac7abd8088719e80f53229bafe
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit d76c8ce876f2609dd9775eb58cfd2ee012b27ad9)

Do not leak Optional in QNameWithPredicateImpl.toString()

The toString() method blindly appends the revision, leading to
Optional[] being present in the result. Fix the method to look
at the returned version and append the query only when it is
present -- just like QName does.

Change-Id: I47dcdaa3c7ca7b3bfaf34898f76b10dfb2ddbe00
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 1951860b7c0ae2a74735a138b3227e09fd718745)

Cleanup LeafRefPath

Remove commented-out code and cleanup javadocs.

Change-Id: I7d6d62bd091191d3ab914a5767eaf4a2ab7830b4
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 8f950a793a451558a137b4057b9680fccf408752)

Consider namespace in namesakes

XML (unlike JSON) has explicit namespaces which we need to take
into consideration when determining whether two elements have the
same name.

JIRA: YANGTOOLS-900
Change-Id: I0d53bee5d18d211343c094941985bd3f0b6a8776
Signed-off-by: Isabel Lloret <illoret@indra.es>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>

Clean various constructs

* Remove intern() on constant strings.
* Avoid some casts by changing test instance types.
* Remove a number of unnecessary type specifiers and casts.
* Remove an unnecessary return.
* Use StringBuilder::append()'s substring parameters instead of a
separate String::substring().

Change-Id: Iaaa206d233fbc43ed1bed2d651c549643e6e7697
Signed-off-by: Stephen Kitt <skitt@redhat.com>
(cherry picked from commit a993ec185764e9f07b7f1c6751bad8f63f6eecc7)

Loosen RpcResultBuilder argument

withRpcErrors can accept a collection of any RpcError subclasses,
express that in the API contract.

Change-Id: I8f302874cb509474e3a974e3b4730338984eda78
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 5bdf2e72626b1cd8c9e98a984e9fc28465d72454)

Do not use SchemaNodeIdentifierBuildNamespace prerequisites

Incremental lookups on ChildSchemaNodeNamespace allow us to react
to childen being created, hence we do not have an assumption that
the target node must be instantiated by someone else (like eager
instantiation).

This will also allow us better recovery if the target node is not
going to appear due to being disabled by if-feature.

JIRA: YANGTOOLS-694
Change-Id: Ibc05e18037bb85d42ed7123293e7f83f896aad5d
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 509988128ed9d5fe314a14906114f53778200e47)

Add NamespaceBehaviour.toString()

A toString() method aids debugging immensely, define some basic
implementations.

Change-Id: I69c6b0027d81ae3a7c6a9304129feef6453e2f1f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 02833cf71885aa15743e44e20f4202269c59de3e)

Omit empty attributes in toString()

Do not emit attributes when they are empty, which brings
the verbosity down quite a bit -- as this is typically the case.

Change-Id: Icb2fa7abb02e9714fed640ec07af1623056a73db
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 032751054483c095a91eb3bf820b133a6b256dc5)

Optimize ImmutableNodes.fromInstanceId()

We are doing a single pass over the YangInstanceIdentifier, hence
we can simply iterate over the nodes, without creating another
YangInstanceIdentifier at each step.

Also cleanup method visibility, mark methods final and merge some
common codepaths.

JIRA: YANGTOOLS-897
Change-Id: I7bd599436dd16f943bae71f3adce289164be25c2
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 38c0fcab1667af91ef8ee0fc1e4b1cd9075f031a)

Fix unique argument parser

Unique argument can use horizontal tabs and line-breaks as separators
in addition to spaces. Deal with them instead of rejecting input.

Change-Id: I5fb6137b9b73c5c844a05cc59fb9261868834a24
JIRA: YANGTOOLS-893
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 68de00c780737c1844f7d045d6f1ecfd0e83f957)

Fix yang-data-jaxen dependencies

We should depend on yang-data-impl for testing only.

Change-Id: Ib7028a1e44f504c08fe5a41707e2089d7d770f27
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit 720fcde7315091a4f867b4aa132b77257d96c9c9)

Bump versions to 2.0.11-SNAPSHOT

This bumps version to next release development.

Change-Id: Ibdef02162752fd1f1cd87fcdbc95e07a83f7f04f
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>