Robert Varga [Wed, 11 Nov 2015 21:34:53 +0000 (22:34 +0100)]
BUG-4629: cleanup YinSchemaSource
This patch introduces Yang/YinSchemaSourceRepresentation, which mark
YANG and YIN schema sources irrespective of the format in which they are
available.
It then adds YimXmlSchemaSource, which allows access to YIN
representation as a javax.xml.transform.Source. Finally it deprecates
YinSchemaSource and provides a utility YinDomSchemaSource, which can
translate a YinXmlSchemaSource in the DOM representation.
Change-Id: I832b2f14637eeb3183c6fe3254f8129bccc0a7ba
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Robert Varga [Fri, 13 Nov 2015 12:50:11 +0000 (13:50 +0100)]
Make a method static
Private method can be made static, make it so.
Change-Id: I21950b4acd92cce20b66a22360285c4c166f519c
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 13 Nov 2015 11:24:18 +0000 (12:24 +0100)]
BUG-865: deprecate DataTreeFactor.create()
This method is provided for compatibility only and should be deprecated,
as the users do not know that they need to specify proper DataTreeType
to get config node enforcement.
Change-Id: I880c75fec2e4a28ae090016229415d0a676a657b
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tony Tkacik [Thu, 12 Nov 2015 13:34:46 +0000 (14:34 +0100)]
Preserve augmentation order in module.
Change-Id: I5a868b7f2fa9a411d66f978784b3fdae2b52fe28
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Filip Gregor [Tue, 10 Nov 2015 13:03:32 +0000 (14:03 +0100)]
Bug 4454: Write to datastore: elements count error
added check for write
added aditional tests
added childrenBefore counter check
Change-Id: I970272520a2c97210cf10a7ff4aa92adc8dc6b27
Signed-off-by: Filip.Gregor <fgregor@cisco.com>
Robert Varga [Tue, 10 Nov 2015 10:02:37 +0000 (11:02 +0100)]
Make InMemorySchemaSourceCache expire items
We use the class only for ASTs. Caching them is good for periods of
parsing churn, such as startup, but the items should expire after not
being used for some time.
Change-Id: I38d95b62fd823906b851466c39429aaab7a72b56
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 9 Nov 2015 22:41:55 +0000 (23:41 +0100)]
Eliminate qname field
The QName always matches the last component of the SchemaPath, hence
eliminate the dedicated field and use SchemaPath#getLastComponent().
Reduces memory footprint of SchemaContext instances by about 1%.
Change-Id: I676b6ef8cd61737b60e7ef7edd86c8210645efd2
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 9 Nov 2015 21:28:38 +0000 (22:28 +0100)]
Split off DeclaredEffectiveStatementBase
Since we are retaining declared statements, which already capture
details from StmtContext, let's specialize a
DesclaredEffectiveStatementBase, which forwards calls to argument and
co. to the declared statement. Reduces memory footprint of
EffectiveSchemaContexts by about 3%.
Change-Id: Icbfa99db841f850c1048fbfe06618613230c2503
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 9 Nov 2015 18:46:03 +0000 (19:46 +0100)]
Share MandatoryStatement instances
Heap dump analysis shows we are wasting memory on MandatoryStatementImpl
instances which are empty. Fix this by creating two dedicated classes
which handle the 'mandatory true' and 'mandatory false' cases.
Same goes for MandatoryEffectiveStatementImpl, for which we provide an
alternative implementation, requiring an MandatoryEffectiveStatement
interface, so that it can be looked up.
Change-Id: If126c3b8d81ad4a870088285ff8956724b199134
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 9 Nov 2015 17:44:52 +0000 (18:44 +0100)]
Lower SchemaPath#getPath() memory overhead
Acquiring forward path involves performing a reverse() on an
ImmutableList -- which results in an proxy object being created.
Each such object costs 24 bytes, leading to ~120K wasted memory on a
freshly-booted up BGP+OF instance.
Eliminate this overhead by instantiating a temporary ArrayList and
copying its reverse view. This way only a single object is retained.
Change-Id: If3b193e84f0f7ffc17997360ae5b02669c119321
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 9 Nov 2015 15:24:57 +0000 (16:24 +0100)]
Instantiate a QName cache
YANG parser needs to suppress duplicate QNames when a grouping is
instantiated to multiple places, for example. This the caching
infrastructure which can be reused.
Change-Id: Ic26298f768cba0a7f76610b93769ae1073c042a2
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 6 Nov 2015 10:44:22 +0000 (11:44 +0100)]
Reuse integer QNames
Use singleton constants instead of instantiating QNames over and over
again.
Change-Id: I1aa94a8c9bfd808d7928589e7f54d8b5224232d0
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Tue, 3 Nov 2015 02:24:33 +0000 (03:24 +0100)]
Cache QNameModules during instantiation
When constructing QNameModules in parser, we want to make sure we reuse
instances as much as possible. To do that, perform cached reference
lookup.
Change-Id: Ibeb26ee89db739a81605913d8fcae6c4e45ed764
Signed-off-by: Robert Varga <rovarga@cisco.com>
Peter Kajsa [Tue, 20 Oct 2015 15:38:13 +0000 (17:38 +0200)]
Bug 4501 - JSON for "type bits" in YANG models not handled correctly
- rejecting of value rewrite (quickfix)
Change-Id: I1007b15d5beacff2877137173fa09e009b877c57
Signed-off-by: Peter Kajsa <pkajsa@cisco.com>
Tom Pantelis [Fri, 2 Oct 2015 16:18:50 +0000 (12:18 -0400)]
Add/improve yang parser error reporting
If a yang file fails to load/parse in karaf, it fails silently. The
first problem is that YangTextSchemaContextResolver#getSchemaContext
logs the SchemaResolutionException to debug as it retries with
getResolvedSources. But if getResolvedSources is empty, it ends up
silently returning an empty SchemaContext.
Another issue is that the underlying exceptions that are really relevant
are added as suppressed to the aggregating exception. This would be fine
if the underlying logger in karaf printed suppressed exceptions as
Throwable#printStackTrace does but it doesn't. The suppressed exceptions
do appear from maven (that underlying logger uses printStackTrace).
My first thought was to log the SchemaResolutionException if
SchemaResolutionException was empty but use printStackTrace to get the
suppressed exceptions. However there are 3 levels of exceptions which is
kind of ugly - one needs to sift thru a lot of noise and long stack
traces to find the relevant error.
So. to make it more user-friendly, I instead modified BundleGlobalContext
to log each source exception's suppressed exceptions as these contain the
actual parsing failures. I also only printed the exception message
normally to avoid the stack trace noise. The full trace can be seen with
debug enabled.
Change-Id: Ie603bab3caa4fabc421ffbd106fddd5176743d2b
Signed-off-by: Tom Pantelis <tpanteli@brocade.com>
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Wed, 4 Nov 2015 21:04:12 +0000 (22:04 +0100)]
BUG-4578: handle APPEARED/DISAPPEARED nodes
These two new events should be handled the same way SUBTREE_MODIFIED is
handled by performing operations on indicated nodes.
Change-Id: I1112f29b91f79712946a5778d1d02666612d38e6
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Tue, 3 Nov 2015 11:48:01 +0000 (12:48 +0100)]
Store key arguments in an ImmutableSet
Make sure we copy key statements from groupings efficiently by reusing
SchemaNodeIdentifiers which have not changed. Also store them in an
ImmutableSet, rather than a HashSet.
Change-Id: I979e417d722bef54e395459403394bdaccdcc339
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 2 Nov 2015 14:42:18 +0000 (15:42 +0100)]
Use an ArrayList instead of LinkedList
ArrayLists are better at append+copy pattern.
Change-Id: I91e6c7344c9622319f0a7859139a14e924e6f2e1
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Tue, 3 Nov 2015 16:13:17 +0000 (17:13 +0100)]
Export yang-model-export via artifacts
This is to enable downstream projects to not define the version of the
dependency.
Change-Id: I0d0861fb5b103abfb475656f1af3ce09ae07b8b1
Signed-off-by: Robert Varga <rovarga@cisco.com>
Maros Marsalek [Tue, 14 Jul 2015 08:43:06 +0000 (10:43 +0200)]
Remove yang-data-operations
Not used anywhere and not needed anymore.
Change-Id: Ie21d18dfaaffe14ddbc843cc568fa1e47004644c
Signed-off-by: Maros Marsalek <mmarsale@cisco.com>
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 2 Nov 2015 14:35:02 +0000 (15:35 +0100)]
Clean YangStatementParserListenerImpl up
Various cleanups:
- unused try/catch block removal
- remove isType local variable
- use Preconditions.checkArgument()
Change-Id: I5a2b15681b9fa8fa3cdaef090a31ea585854da7e
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 2 Nov 2015 14:28:13 +0000 (15:28 +0100)]
Switch StatementContextBase to use ArrayList
This inlines the history initialization and uses an ArrayList instead of
a LinkedList.
Change-Id: I5980e6525f0b00c46b1f1f87dad941685042abe9
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 2 Nov 2015 12:19:23 +0000 (13:19 +0100)]
Do not use String.replace()
Change-Id: I2ad218ca9a297300a506ec5984ab1b2fc4d86efa
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Mon, 2 Nov 2015 14:21:19 +0000 (15:21 +0100)]
Remove a search-before-remove on ArrayList
There is no need to make the removal conditional on presence. This
forces two linear searches of the list, instead of just one.
Change-Id: I3d3fb23ac02e95433cc17ccee119dbb1392cef42
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sun, 1 Nov 2015 18:27:29 +0000 (19:27 +0100)]
Share empty ConstraintDefinition instances
Introduce a specialized class with two singleton instances for holding
the (very common) empty ConstraintDefinition object. Hide
EffectiveConstraintDefinitionImpl and make it expose just a static
factory method, which does the right thing.
Also creates ConstraintDefinitions to hold common functions for the two
implementations.
Change-Id: I4c25aee638ddb0617030fffcb0bc54ac13d28bc1
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sun, 1 Nov 2015 17:08:27 +0000 (18:08 +0100)]
Make SchemaNodeIdentifier.Relative final
This makes sure we do not ever subclass it, preventing implementation
leakage.
Change-Id: I8901600968db777342ebb858279cb69b3e35997c
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sun, 1 Nov 2015 13:59:41 +0000 (14:59 +0100)]
Use ImmutableList for unknown nodes
Do not allow modification of unknown nodes, also saves some 30KiB from
1200 empty ArrayLists.
Change-Id: Ie0fa44ebdca237a31928718445e2b2deb41a2299
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sun, 1 Nov 2015 13:54:35 +0000 (14:54 +0100)]
Share Integer.MAX_VALUE
As it turns out Integer.MAX_VALUE is not usually cached for boxing. Add
an explicit cache to eliminate ~160K Integers wasting 2.5MB on a simple
SchemaContext.
Change-Id: I47b5b00358d5726eddc8baf675bda1f850823362
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sun, 1 Nov 2015 12:09:02 +0000 (13:09 +0100)]
BUG-4556: enable DataSchemaContextTrees to be scavenged
DataSchemaContextTree contains an internal weak-keyed cache, which is
used to speed up lookups and share DataSchemaContextNodes.
Unfortunately the cache uses strong references for values (e.g.
DataSchemaContextTree) instances, which transitively (via
DataSchemaContextNode) retain the reference to the key. This has the
effect of the key being always reachable, thus causing a memory leak.
Make the cache use weak references for values, which ensures that it
gets scavenged whenever its contents are no longer referenced.
Change-Id: I1684d1bf2c2c5cdc44cda1e8e508149ff3a8fae4
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sat, 31 Oct 2015 13:10:16 +0000 (14:10 +0100)]
BUG-4556: Introduce StmtContext.getSchemaPath()
SchemaContexts are created in top-to-bottom direction, making them
useful for holding SchemaPaths. This reduces the memory overhead
from not constructing SchemaPaths in a hierarchical manner.
Change-Id: I85a49c6c861543a585730c8dac12622605eb20d5
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Martin Ciglan [Fri, 30 Oct 2015 14:27:42 +0000 (15:27 +0100)]
Capture ANTLR4 syntax errors and log them
By default, ANTLR4 parser instance has ANTLR4's ConsoleErrorListener
attached to it, therefore error messages are piped to System.err.
In this patch, this is silenced and captured via existing
custom implementation of ANTLR4's BaseErrorListener.
Change-Id: I1922880bf09e69ca55316dacbca9b5fc338f721c
Signed-off-by: Martin Ciglan <mciglan@cisco.com>
Peter Kajsa [Wed, 28 Oct 2015 15:04:03 +0000 (16:04 +0100)]
Bug 4504 - YIN generation failed on openflow-protocol module
Unknown nodes added by augmentation shouldn't be exported by yin export
of a particular yang node.
Change-Id: I421aeee7732f40929f902546384c0a7bab0bf22c
Signed-off-by: Peter Kajsa <pkajsa@cisco.com>
Robert Varga [Sat, 24 Oct 2015 18:18:59 +0000 (20:18 +0200)]
Make TestModel a utility class
Final and private contructor.
Change-Id: Ia42bda3b878caa44aa785529d38928f167a863fc
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Thu, 22 Oct 2015 20:39:39 +0000 (22:39 +0200)]
Do not leak internal methods
applyTouch() is method internal to implementation, it should be left
protected, not public.
Change-Id: Ic57e5bffe66171747650bf13f501f89f8a59fd57
Signed-off-by: Robert Varga <rovarga@cisco.com>
Jan Hajnar [Wed, 20 May 2015 09:47:30 +0000 (11:47 +0200)]
BUG-2399: Implement automatic container removal
Structural containers all all container nodes which do not contain a
presence statement. RFC6020 specifies that structural containers have no
semantic meaning aside from being containment nodes for their children
and therefore can be created when a child would appear and should
disappear when they lose their last child.
Containers which should not be subject to this lifecycle need to be
marked with a 'presence' statement, which attaches semantic meaning
to their presence.
Change-Id: Icb69c959a27726b549f563b84b52727592b67cb1
Signed-off-by: Robert Varga <rovarga@cisco.com>
Signed-off-by: Jan Hajnar <jhajnar@cisco.com>
Martin Ciglan [Mon, 12 Oct 2015 07:40:46 +0000 (09:40 +0200)]
Bug 4414: Shorthand case statement not working
Original for shorthand-case was not set at all,
therefore definition in grouping causes
IncorrectNestingException: Supplied class is not valid case
- code refactored
- fix: getOriginal() returns proper value
Change-Id: I6941eea52a07967824d2653fbfaedebc5f35ddc9
Signed-off-by: Martin Ciglan <mciglan@cisco.com>
Robert Varga [Fri, 23 Oct 2015 08:51:29 +0000 (10:51 +0200)]
Lists with min-elements > 0 are mandatory
As per https://tools.ietf.org/html/rfc6020#section-3.1, leaf-list and
list items which have min-elements greater than zero are considered
mandatory.
Adjust ConstrainDefinition implementations to take min-elements into
account and report isMandatory() as true in case it is non-zero.
Change-Id: I7d7a0c15ce26bab2eed5bc2569cc6b8904dac203
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tony Tkacik [Thu, 22 Oct 2015 09:01:49 +0000 (11:01 +0200)]
Do not retain build context via extensions.
Change-Id: I3a074d5c526668e85a516678f09b419f4d335347
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Robert Varga [Sun, 18 Oct 2015 11:20:15 +0000 (13:20 +0200)]
Add range validation and normalization
RangeRestrictedTypeBuilder now computes proper ranges, lowered to the
class defined by the base type, so we have the same types for
enforcement no matter the input.
Change-Id: Ibc139a3ff3d18bba87219a1fe91b2269b107a3c6
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Robert Varga [Sun, 18 Oct 2015 11:19:52 +0000 (13:19 +0200)]
Fix various small warnings
Unused import and raw type reference.
Change-Id: I625813261450ff85dc349da19bfbfceab80ff305
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Robert Varga [Mon, 19 Oct 2015 13:11:02 +0000 (15:11 +0200)]
Introduce UnresolvedNumber
UnresolvedNumber is a better-designed replacement for
UnknownBoundaryNumber.
Change-Id: Id603529d9469684d853080d702db50637afe67b3
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 16 Oct 2015 17:52:10 +0000 (19:52 +0200)]
Util types: implement hashCode()/equals()/toString()
This patch adds the infrastructure and provides the base
hashCode/equals/toString implementations for all types.
Change-Id: I8e10bcc71e41bc07e5ff96f2746e3352b02abc39
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Robert Varga [Fri, 16 Oct 2015 12:20:34 +0000 (14:20 +0200)]
Do not lookup root operation explicitly
Make sure we use the public API and not rely of internal workings in
tests.
Change-Id: I435a385e8279a1d1fb4f4860767c96846f759880
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Robert Varga [Fri, 16 Oct 2015 12:42:18 +0000 (14:42 +0200)]
Reuse no-revision qnames
The qname can be used twice, it makes sense to keep it around.
Change-Id: Iedf0fffb7a0791bb264edb43cb734d4e1289955e
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 16 Oct 2015 13:03:22 +0000 (15:03 +0200)]
BUG-4261: intern leaf nodes
Look up the interner for leaf nodes based on the current schema node.
Change-Id: Ibd7ebc3affa315c650c3dce1cae7138d9aed43ab
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Robert Varga [Sat, 10 Oct 2015 21:18:07 +0000 (23:18 +0200)]
Introduce model.util.type package
This package introduces new utility classes for implementing a
TypeDefinition hierarchy. The two entrypoints are RestrictedTypes and
DerivedTypes.
Change-Id: Icf0a91f09151c975a77461ee4488af1f5b300d27
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Robert Varga [Thu, 15 Oct 2015 02:18:37 +0000 (04:18 +0200)]
Introduce TypeDefinitionAware
TypeEffectiveStatement and TypedefEffectiveStatement should expose the
underlying TypeDefinition, so it can be acquired for use with the type
system.
Change-Id: I58aaa33df077fc092db35f87be671a0b89e50fd3
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Igor Foltin [Tue, 29 Sep 2015 08:49:25 +0000 (10:49 +0200)]
Bug 3899: Milestone: Increase test coverage for YangTools (80%).
- fixed wildcard imports and missing headers in test classes
Change-Id: Iaf12398e8190089daff248533bc4af1cc82a829e
Signed-off-by: Igor Foltin <igor.foltin@pantheon.sk>
Signed-off-by: Igor Foltin <ifoltin@cisco.com>
Peter Kajsa [Mon, 7 Sep 2015 15:12:35 +0000 (17:12 +0200)]
Bug 3799: NPE when including a submodule of a more recent revision than the module
Change-Id: Ieaff77303879d9f9eee4f4ae4840a517ada66a91
Signed-off-by: Peter Kajsa <pkajsa@cisco.com>
Robert Varga [Sat, 26 Sep 2015 12:38:28 +0000 (14:38 +0200)]
Cleanup a few warnings
Import ordering, methods which can be made static.
Change-Id: I0448e21fde4d54f264429ebbd0928b408d29af65
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sun, 4 Oct 2015 22:21:22 +0000 (00:21 +0200)]
Remove unused imports
Just a drive-by cleanup job.
Change-Id: I2b6620ee970e1bdaf0c2f99a61f089f46817c03f
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Tue, 29 Sep 2015 05:34:48 +0000 (07:34 +0200)]
LeafRefUtils is a utility class
Should be final with a private constructor.
Change-Id: I5c45da74cc3cefc51094e3663beba690c08eef9d
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tony Tkacik [Tue, 13 Oct 2015 08:13:58 +0000 (10:13 +0200)]
Bug 4349: Allowed UnkeyedListItem to be start event in Writer.
Change-Id: I42fd224974e6fcc691c548978beebce9a6828257
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Filip Gregor [Fri, 9 Oct 2015 12:43:44 +0000 (14:43 +0200)]
Bug 4412: New yang parser effective statements cleanup
- improved classes hierarchy and reduced source code duplicity
- fixed some major sonar issues
Change-Id: Ibec3da63b0a3ab0616402b52798e9450df7feb70
Signed-off-by: Peter Kajsa <pkajsa@cisco.com>
Signed-off-by: Filip Gregor <filip.gregor@pantheon.sk>
Robert Varga [Sat, 10 Oct 2015 18:05:18 +0000 (20:05 +0200)]
BUG-4322: Leafref should not have a default value
RFC6020 does not mention anything about the default value for leaf
references.
Change-Id: I0edcfe286aa6ac6fd93bb5f0074a51aeb18627d8
Signed-off-by: Robert Varga <nite@hq.sk>
Robert Varga [Fri, 9 Oct 2015 17:14:51 +0000 (19:14 +0200)]
BUG-4261: convert JSON parser to pass down SchemaNode information
Change-Id: I2519e3b77d88394c425e79610208a7c8d4a98b9f
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 9 Oct 2015 17:56:47 +0000 (19:56 +0200)]
BUG-4261: make ImmutableNormalizedNodeStreamWriter DataSchemaNodeAware
ImmutableNormalizedNodeStreamWriter can perform leaf node interning when
it knows what the schema is. Turn it into a DataSchemaNodeAware, so
users can pass down this information and make use of it when adding
leaves.
Change-Id: I7566a8705c0a05ea3cfa1335c7335021de6edccc
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 9 Oct 2015 16:31:02 +0000 (18:31 +0200)]
BUG-4261: introduce DataSchemaNodeAware NormalizedNodeStreamWriter
This variant allows passing a DataSchemaNode from the user.
Implementations can take advantage of this information to optimize their
result. Utility no-op adaptors are provided, too.
Change-Id: I990197baf3a80f9826cd35d3bb3155d1b7a10cc3
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sat, 10 Oct 2015 11:46:55 +0000 (13:46 +0200)]
BUG-4261; split NormalizedNodeInterners
As it turns out, having separate classes is easier for users.
LeafsetEntryInterner can be used to check the type once and then rely on
this check. LeafInterner returns a no-op instance for convenience if the
schema is not available.
Change-Id: Iec2117be2f91114943d2896a6bfb273f683ea042
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 9 Oct 2015 16:59:31 +0000 (18:59 +0200)]
Use StringBuilder instead of StringWriter
We do not need a Writer interface, as we are simply compositing a
String.
Change-Id: Ied863fe0bf39e99dffe83c2520c56e735d4102d4
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 9 Oct 2015 16:31:27 +0000 (18:31 +0200)]
Suppress Version.nextVersion() warning
The method could be static based on the currect implementation, but we
want to keep it an instance-method.
Change-Id: Ibb604f1038e6df9a6731504af67dc7d960dee6bf
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sat, 10 Oct 2015 13:47:30 +0000 (15:47 +0200)]
BUG-4407: Fix a typo in Decimal64 range tables
Simple typo results in both wrong range enforcement and an out-of-bounds
error with fraction-digits == 18.
Change-Id: I3e9916940e80fbeedadfdc9e8db35b584ea3e859
Signed-off-by: Robert Varga <robert.varga@pantheon.sk>
Igor Foltin [Tue, 6 Oct 2015 11:43:45 +0000 (13:43 +0200)]
Bug 4314: maven plugin: switch logging
to slf4j
Cleaned up use of maven Log in
yang-maven-plugin and switched to
slf4j.
Cleaned up use of static method
Util.message.
Change-Id: I6bf77b761facabfec5ba3a64b193e41b4dfc870f
Signed-off-by: Igor Foltin <igor.foltin@pantheon.sk>
Signed-off-by: Igor Foltin <ifoltin@cisco.com>
Filip Gregor [Thu, 8 Oct 2015 08:31:10 +0000 (10:31 +0200)]
Bug 3899: Increase test coverage for YangTools
yang-parser rfc6020 effective type test coverage
Change-Id: I2cfc8eab9d52070907d53a8dacd376ffa2a1a170
Signed-off-by: Filip Gregor <filip.gregor@pantheon.sk>
Robert Varga [Mon, 5 Oct 2015 17:30:13 +0000 (19:30 +0200)]
BUG-4359: use WRITE-implied metadata
When we apply an operation under a WRITE node, we need to look up the
metadata based on that write to correctly understand what the effect of
this operation is.
This is important when we modify data introduced by the WRITE node, as
without this lookup a WRITE/DELETE combination on the child will result
in a NONE logical operation, which means data introduced in the WRITE
node will not be modified and seemingly reappear.
Change-Id: I5a5a67470f1f8baafc7a71069341dead073ba84a
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sun, 4 Oct 2015 02:32:00 +0000 (04:32 +0200)]
BUG-4407: Fix Decimal64 range enforcement
Baseline ranges did not take into account the number of fraction digits,
but rather instantiated the most permissive range. Fix this by
pre-calculating all ranges specified in RFC6020 Section 9.3.4.
Also enforce SchemaPath presence, as that is the unique identifier of
the type.
Change-Id: I5e8799da618398ba458606226e71e434f5214f33
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tony Tkacik [Tue, 6 Oct 2015 13:55:25 +0000 (15:55 +0200)]
Speed-up look up of namespace listeners.
Change-Id: Ie704318ddabc94cf87ee7902c0185812e721da43
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Tony Tkacik [Mon, 5 Oct 2015 08:42:27 +0000 (10:42 +0200)]
Fixed shared state in stateless ModuleStatementSupport.
Change-Id: I613e71b61038e7354db6a34559b5745baaeb7d09
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Tony Tkacik [Sun, 4 Oct 2015 16:36:32 +0000 (18:36 +0200)]
Speeded up ANTLR tree walk for statement parser.
Change-Id: I49b28a3cb97bb2bb4f0f133d2ed6c97293cf59bd
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Robert Varga [Sun, 4 Oct 2015 05:14:55 +0000 (07:14 +0200)]
Deprecate InstanceIdentifierTypeDefinition.getRevisionAwareXPath()
This method has been added due to mis-reading of RFC6020. YANG actually
does not define this property, but rather specifies how
instance-identifier types are to be interpreted.
Change-Id: I484b655475f2817277b62dae5fc7462442ebf874
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 2 Oct 2015 22:43:19 +0000 (00:43 +0200)]
Use a proper NamespaceContext for XPaths
Parsing an XPath which contains namespace prefixes requires a
NamespaceContext. Create a StmtContext-backed implementation and pass it
to XPath before we attempt to parse an XPath string.
Change-Id: Ib28cac2525ffcb6b8e3026e19e431cb9dc86c14b
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tony Tkacik [Sat, 3 Oct 2015 12:22:49 +0000 (12:22 +0000)]
Revert "Revert "Updated SchemaNodeIdentifier namespace handling.""
Bringing back updated augment lookup and validation code
which turned out to be correct, but unfortunatelly original
implementation of parser (from Hydrogen) accepted incorrect
values and "normalized" them to probably what user really intented.
New code is stricter and did not accept that input, since
it really was unclear what user intented.
Change-Id: I187c3479fd7bbb986f414805401e8b54f6b49ee0
Release-Note: Incompatible-Behaviour
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Robert Varga [Sat, 3 Oct 2015 01:54:57 +0000 (03:54 +0200)]
Centralize RevisionAwareXPath creation
Three call sites are performing the same thing, create a common method.
Change-Id: I274f23b1785210cef96465e4db944c3b923886a8
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sat, 3 Oct 2015 01:41:35 +0000 (03:41 +0200)]
Do not compile augment/refine/deviate paths
These paths are not full XPaths, but rather SchemaNodeIdentifier paths.
Do not attempt to compile them, as we validate them by splitting and
look up of individual components.
Change-Id: If472ff6a2c795ce810794ad64dc643bf10a24199
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Sat, 3 Oct 2015 00:23:18 +0000 (02:23 +0200)]
Use a ThreadLocal XPathFactory
XPathFactory is documented to be non-threadsafe. While we could use
single shared instance, accessing it would require synchronization. Use
a ThreadLocal, so each thread gets its own instance.
Change-Id: I3f6a722df94d261c0680b30b9b65b73566350794
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 2 Oct 2015 23:36:12 +0000 (01:36 +0200)]
Split keys into an ImmutableSet
This simplifies building the set and results in memory-optimal set.
Change-Id: I16cde5188b7c473c8cac2943e587ff553cb31c05
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 2 Oct 2015 23:06:22 +0000 (01:06 +0200)]
Expose LIST_KEY_SPLITTER
LIST_KEY_SEPARATOR is only ever used to instantiate a splitter, which is
what we already have a constant for. Remove LIST_KEY_SEPARATOR and make
its external user call LIST_KEY_SPLITTER, speeding it up.
Change-Id: I7144bd07903be4b2982f5341c5cdeb6956af5187
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 2 Oct 2015 22:55:46 +0000 (00:55 +0200)]
Add @SafeVargs
findFirstDeclaredSubstatement() gets an array of type-restricted Classes
and iterates over them. This is safe usage, hence the anotation is
appropriate.
Change-Id: Ie57f191623e723a34b76e8ea5fe39e3b04df4cb2
Signed-off-by: Robert Varga <rovarga@cisco.com>
Filip Gregor [Fri, 2 Oct 2015 13:28:21 +0000 (15:28 +0200)]
Bug 3876: Introduce support for XPath / XQuery queries on YANG Data
- unit test coverage increase 60%
- fixed broken xpath with predicates
Change-Id: I853772a28370cdf5901ec8b01ca3ad1d105bb854
Signed-off-by: Filip Gregor <filip.gregor@pantheon.sk>
Tony Tkacik [Fri, 2 Oct 2015 22:33:46 +0000 (22:33 +0000)]
Revert "Updated SchemaNodeIdentifier namespace handling."
This reverts commit
b6e9bb9ff439eb0e905d8ccf61e15a5a3b2bf444.
Change-Id: I927dc1a9973ab36e7bc54c823e93fd00b240d123
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Tony Tkacik [Thu, 1 Oct 2015 17:23:11 +0000 (19:23 +0200)]
Updated SchemaNodeIdentifier namespace handling.
Introduced notion of derived namespaces and SchemaNodeIdentifier
namespace for lookups of refine / augment and other statements
- Fixed AugmentUtils and AugmentStatementImpl, UsesStatementImpl
to use declarative way to get needed artefacts instead of
manual search in statement context.
- Removed utility methods, which are not needed anymore.
Change-Id: Ifb9c6dffa948cf265f81f2d8cd8615318b1447d0
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Stephen Kitt [Fri, 18 Sep 2015 16:15:35 +0000 (18:15 +0200)]
Avoid NPEs in LoggerMustBeSlf4jCheck
aAST.getFirstChild().findFirstToken(TokenTypes.IDENT) is often null in
ODL builds, so check it before retrieving the text.
Change-Id: Ic9905f72d1b2ac9bd902e8eb30b8f48f1b8930d2
Signed-off-by: Stephen Kitt <skitt@redhat.com>
Robert Varga [Fri, 2 Oct 2015 11:33:25 +0000 (13:33 +0200)]
BUG-865: Deprecate ExtendedType
This thing has been replaced with the more semantically correct (and
less confusing) DerivedType.
Change-Id: I4491a9c47f43b18a7d4d8a7edc08e201856ec7cc
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 2 Oct 2015 16:32:12 +0000 (18:32 +0200)]
BUG-4400: make sure we retain input ordering
For key statements we must retain the definition order, so use a
LinkedHashSet instead of a regular HashSet.
Change-Id: I9d327e571b895ee4b9dadeb5c5d0817e81a4906c
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tony Tkacik [Fri, 2 Oct 2015 12:38:04 +0000 (14:38 +0200)]
Fixed ordering of keys in key statement.
Change-Id: I2a02ce7f9507ae292dac8b2173d3fb24e5fe00e5
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Robert Varga [Thu, 1 Oct 2015 13:46:47 +0000 (15:46 +0200)]
BaseTypes should reference into YangConstants
We should not be duplicating namespace/module definitions.
Change-Id: Idc76a9caae2b78112b3c6ead4ac6772ea4d13222
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Thu, 1 Oct 2015 11:13:48 +0000 (13:13 +0200)]
Use shared Splitter instance
Do not instantiate a temporary splitter, but reuse the global one.
Change-Id: Ida0d6a725aab4731cd787ca6719e33c002422b5e
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tony Tkacik [Fri, 2 Oct 2015 09:28:57 +0000 (11:28 +0200)]
Fixed allModuleIdentifiers
allModuleIdentifiers was incorrectly populated from optional
part of API instead of required part of API,
which leaded to misrepporting identifiers to users.
Change-Id: I6e2d4b84137211fef1446b047cb6e5d3cea3101f
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Martin Ciglan [Thu, 1 Oct 2015 09:33:20 +0000 (11:33 +0200)]
General issue: In case of uncompilable XPathExpression,
log warning rather than throwing exception.
Change-Id: I16b45cf57b9cf57e21221101cf66335059718982
Signed-off-by: Martin Ciglan <mciglan@cisco.com>
Robert Varga [Thu, 1 Oct 2015 11:07:24 +0000 (13:07 +0200)]
Use a pre-compiled pattern in SchemaContextUtil
String.replaceAll() performs an implicit compilation -- use a
pre-compiled pattern instead.
Change-Id: I5323d19d8b6c009d837a14e32d91ba711cd4d486
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Thu, 1 Oct 2015 10:50:56 +0000 (12:50 +0200)]
Use explicit ImmutableList instead of FluentIterable
FluentIterable.transform() creates a temporary object which we do not
need -- just perform an ecplicit copy.
Change-Id: Ie599c1b0b7c4eb81fbfd5ab8a5a4c16389cb2fb7
Signed-off-by: Robert Varga <rovarga@cisco.com>
Thanh Ha [Wed, 30 Sep 2015 21:50:25 +0000 (17:50 -0400)]
Use odlparent-lite for artifacts pom
Change-Id: I663bd08a28c4e7abc39e5b8eef8b9f1ab6664bf9
Signed-off-by: Thanh Ha <thanh.ha@linuxfoundation.org>
Robert Varga [Fri, 18 Sep 2015 09:39:42 +0000 (11:39 +0200)]
Cleanup AugmentUtils
Du no copy collections, user preconditions, use pre-instantiated sets
and similar.
Change-Id: I1fadd465650323504515e3df9393d42c8593c1e8
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Thu, 1 Oct 2015 06:44:16 +0000 (08:44 +0200)]
Cleanup imports
Unused imports should be removed. Also remove unused isMap local
variable.
Change-Id: I4c45aebe0538f84b87828ac9725cfe121d3befa3
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Thu, 1 Oct 2015 06:43:20 +0000 (08:43 +0200)]
Cleanup TypeUtils
Preconditions, Immutable{Map,Set}, comparator speed up.
Change-Id: Ib4a1673aa0c3bf11ec7c69b3b3fcf1352d7c4e56
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tom Pantelis [Thu, 1 Oct 2015 07:11:28 +0000 (03:11 -0400)]
Fix performance issue in EffectiveStatementBase
While running some feature tests I took some thread dumps and noticed
the following stack trace in each:
java.lang.Thread.State: RUNNABLE
at java.lang.Throwable.fillInStackTrace(Native Method)
at java.lang.Throwable.fillInStackTrace(Throwable.java:783)
- locked <0x00000000e5aff880> (a
java.util.NoSuchElementException)
at java.lang.Throwable.<init>(Throwable.java:250)
at java.lang.Exception.<init>(Exception.java:41)
at java.lang.RuntimeException.<init>(RuntimeException.java:51)
at
java.util.NoSuchElementException.<init>(NoSuchElementException.java:47)
at
com.google.common.collect.AbstractIterator.next(AbstractIterator.java:154)
at com.google.common.collect.Iterators.find(Iterators.java:717)
at com.google.common.collect.Iterables.find(Iterables.java:646)
at
org.opendaylight.yangtools.yang.parser.stmt.rfc6020.effective.EffectiveStatementBase.firstEffective(EffectiveStatementBase.java:105)
The problem is that Iterables.find throws NoSuchElementException if not
found which is expensive when called many times. I changed it to call
tryFind instead which returns an Optional and this significantly
improved performance. The test time went from 15 min to 3 min.
There may be other classes that call Iterables.find - I didn't check.
If so they can be fixed in subsequent patches.
Change-Id: I0b3d02979b6d7fc97752c4b2429720b0807b853b
Signed-off-by: Tom Pantelis <tpanteli@brocade.com>
Robert Varga [Thu, 28 May 2015 11:43:12 +0000 (13:43 +0200)]
Enforce SchemaPath correctness
SchemaPath must never contain a null element unless it is ROOT or SAME.
Current implementation allows that to happen by checking for null items
in the collections/lists it gets from the user.
Change-Id: Ibb85f84f54ad9c9049ac2ce8bb88371942b16880
Signed-off-by: Robert Varga <rovarga@cisco.com>
Robert Varga [Fri, 18 Sep 2015 09:52:45 +0000 (11:52 +0200)]
Cleanup yang-parser-impl Utils
Pre-compiled paterns, reused splitters and similar.
Change-Id: Ia8b6aebb0d3d1922018078b3e33ebba7e9740769
Signed-off-by: Robert Varga <rovarga@cisco.com>
Tony Tkacik [Wed, 30 Sep 2015 20:00:50 +0000 (22:00 +0200)]
Introduced ChildSchemaNodes namespace to parser
Use of purpose-built namespaces in new parser
simplifies code and speed ups search.
Introduced ChildSchemaNodes namespace in order
to speed up search for refines and uses.
Change-Id: Ie716b5798a3cca77dff3e3799383f82649273f97
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>
Tony Tkacik [Wed, 30 Sep 2015 16:56:43 +0000 (18:56 +0200)]
Fixed parser failing to continue parsing on '\n\n \n\n '"
Statement grammar used hidden tokens to skip comments
which in some cases when there was emtpy line, comment, empty
line ended up emitting two SEP tokens, but parser rules
were written expecting only zero or one SEP tokens.
Changing parser rules to expect multiple SEP tokens,
fixed that issue.
Change-Id: Ic3ea43f520582863681740534aa42739069056a3
Signed-off-by: Tony Tkacik <ttkacik@cisco.com>