Class RevWalk

  • All Implemented Interfaces:
    AutoCloseable, Iterable<RevCommit>
    Direct Known Subclasses:
    DepthWalk.RevWalk, ObjectWalk, PlotWalk

    public class RevWalk
    extends Object
    implements Iterable<RevCommit>, AutoCloseable
    Walks a commit graph and produces the matching commits in order.

    A RevWalk instance can only be used once to generate results. Running a second time requires creating a new RevWalk instance, or invoking reset() before starting again. Resetting an existing instance may be faster for some applications as commit body parsing can be avoided on the later invocations.

    RevWalk instances are not thread-safe. Applications must either restrict usage of a RevWalk instance to a single thread, or implement their own synchronization at a higher level.

    Multiple simultaneous RevWalk instances per Repository are permitted, even from concurrent threads. Equality of RevCommits from two different RevWalk instances is never true, even if their ObjectIds are equal (and thus they describe the same commit).

    The offered iterator is over the list of RevCommits described by the configuration of this instance. Applications should restrict themselves to using either the provided Iterator or next(), but never use both on the same RevWalk at the same time. The Iterator may buffer RevCommits, while next() does not.

    • Constructor Detail

      • RevWalk

        public RevWalk​(Repository repo)
        Create a new revision walker for a given repository.
        Parameters:
        repo - the repository the walker will obtain data from. An ObjectReader will be created by the walker, and will be closed when the walker is closed.
      • RevWalk

        public RevWalk​(ObjectReader or)
        Create a new revision walker for a given repository.

        Parameters:
        or - the reader the walker will obtain data from. The reader is not closed when the walker is closed (but is closed by dispose().
    • Method Detail

      • getObjectReader

        public ObjectReader getObjectReader()
        Get the reader this walker is using to load objects.
        Returns:
        the reader this walker is using to load objects.
      • createReachabilityChecker

        @Deprecated
        public final ReachabilityChecker createReachabilityChecker()
                                                            throws IOException
        Deprecated.
        use ObjectReader#createReachabilityChecker(RevWalk) instead.
        Get a reachability checker for commits over this revwalk.
        Returns:
        the most efficient reachability checker for this repository.
        Throws:
        IOException - if it cannot open any of the underlying indices.
        Since:
        5.4
      • close

        public void close()

        Release any resources used by this walker's reader.

        A walker that has been released can be used again, but may need to be released after the subsequent usage.

        Specified by:
        close in interface AutoCloseable
        Since:
        4.0
      • markStart

        public void markStart​(RevCommit c)
                       throws MissingObjectException,
                              IncorrectObjectTypeException,
                              IOException
        Mark a commit to start graph traversal from.

        Callers are encouraged to use parseCommit(AnyObjectId) to obtain the commit reference, rather than lookupCommit(AnyObjectId), as this method requires the commit to be parsed before it can be added as a root for the traversal.

        The method will automatically parse an unparsed commit, but error handling may be more difficult for the application to explain why a RevCommit is not actually a commit. The object pool of this walker would also be 'poisoned' by the non-commit RevCommit.

        Parameters:
        c - the commit to start traversing from. The commit passed must be from this same revision walker.
        Throws:
        MissingObjectException - the commit supplied is not available from the object database. This usually indicates the supplied commit is invalid, but the reference was constructed during an earlier invocation to lookupCommit(AnyObjectId).
        IncorrectObjectTypeException - the object was not parsed yet and it was discovered during parsing that it is not actually a commit. This usually indicates the caller supplied a non-commit SHA-1 to lookupCommit(AnyObjectId).
        IOException - a pack file or loose object could not be read.
      • markUninteresting

        public void markUninteresting​(RevCommit c)
                               throws MissingObjectException,
                                      IncorrectObjectTypeException,
                                      IOException
        Mark a commit to not produce in the output.

        Uninteresting commits denote not just themselves but also their entire ancestry chain, back until the merge base of an uninteresting commit and an otherwise interesting commit.

        Callers are encouraged to use parseCommit(AnyObjectId) to obtain the commit reference, rather than lookupCommit(AnyObjectId), as this method requires the commit to be parsed before it can be added as a root for the traversal.

        The method will automatically parse an unparsed commit, but error handling may be more difficult for the application to explain why a RevCommit is not actually a commit. The object pool of this walker would also be 'poisoned' by the non-commit RevCommit.

        Parameters:
        c - the commit to start traversing from. The commit passed must be from this same revision walker.
        Throws:
        MissingObjectException - the commit supplied is not available from the object database. This usually indicates the supplied commit is invalid, but the reference was constructed during an earlier invocation to lookupCommit(AnyObjectId).
        IncorrectObjectTypeException - the object was not parsed yet and it was discovered during parsing that it is not actually a commit. This usually indicates the caller supplied a non-commit SHA-1 to lookupCommit(AnyObjectId).
        IOException - a pack file or loose object could not be read.
      • isMergedInto

        public boolean isMergedInto​(RevCommit base,
                                    RevCommit tip)
                             throws MissingObjectException,
                                    IncorrectObjectTypeException,
                                    IOException
        Determine if a commit is reachable from another commit.

        A commit base is an ancestor of tip if we can find a path of commits that leads from tip and ends at base.

        This utility function resets the walker, inserts the two supplied commits, and then executes a walk until an answer can be obtained. Currently allocated RevFlags that have been added to RevCommit instances will be retained through the reset.

        Parameters:
        base - commit the caller thinks is reachable from tip.
        tip - commit to start iteration from, and which is most likely a descendant (child) of base.
        Returns:
        true if there is a path directly from tip to base (and thus base is fully merged into tip); false otherwise.
        Throws:
        MissingObjectException - one or more of the next commit's parents are not available from the object database, but were thought to be candidates for traversal. This usually indicates a broken link.
        IncorrectObjectTypeException - one or more of the next commit's parents are not actually commit objects.
        IOException - a pack file or loose object could not be read.
      • getMergedInto

        public List<Ref> getMergedInto​(RevCommit commit,
                                       Collection<Ref> refs)
                                throws IOException
        Determine the Refs into which a commit is merged.

        A commit is merged into a ref if we can find a path of commits that leads from that specific ref and ends at commit.

        Parameters:
        commit - commit the caller thinks is reachable from refs.
        refs - refs to start iteration from, and which is most likely a descendant (child) of commit.
        Returns:
        list of refs that are reachable from commit.
        Throws:
        IOException - a pack file or loose object could not be read.
        Since:
        5.12
      • getMergedInto

        public List<Ref> getMergedInto​(RevCommit commit,
                                       Collection<Ref> refs,
                                       ProgressMonitor monitor)
                                throws IOException
        Determine the Refs into which a commit is merged.

        A commit is merged into a ref if we can find a path of commits that leads from that specific ref and ends at commit.

        Parameters:
        commit - commit the caller thinks is reachable from refs.
        refs - refs to start iteration from, and which is most likely a descendant (child) of commit.
        monitor - the callback for progress and cancellation
        Returns:
        list of refs that are reachable from commit.
        Throws:
        IOException - a pack file or loose object could not be read.
        Since:
        5.12
      • isMergedIntoAny

        public boolean isMergedIntoAny​(RevCommit commit,
                                       Collection<Ref> refs)
                                throws IOException
        Determine if a commit is merged into any of the given refs.
        Parameters:
        commit - commit the caller thinks is reachable from refs.
        refs - refs to start iteration from, and which is most likely a descendant (child) of commit.
        Returns:
        true if commit is merged into any of the refs; false otherwise.
        Throws:
        IOException - a pack file or loose object could not be read.
        Since:
        5.12
      • isMergedIntoAll

        public boolean isMergedIntoAll​(RevCommit commit,
                                       Collection<Ref> refs)
                                throws IOException
        Determine if a commit is merged into all of the given refs.
        Parameters:
        commit - commit the caller thinks is reachable from refs.
        refs - refs to start iteration from, and which is most likely a descendant (child) of commit.
        Returns:
        true if commit is merged into all of the refs; false otherwise.
        Throws:
        IOException - a pack file or loose object could not be read.
        Since:
        5.12
      • getRevSort

        public EnumSet<RevSort> getRevSort()
        Obtain the sort types applied to the commits returned.
        Returns:
        the sorting strategies employed. At least one strategy is always used, but that strategy may be RevSort.NONE.
      • hasRevSort

        public boolean hasRevSort​(RevSort sort)
        Check whether the provided sorting strategy is enabled.
        Parameters:
        sort - a sorting strategy to look for.
        Returns:
        true if this strategy is enabled, false otherwise
      • sort

        public void sort​(RevSort s)
        Select a single sorting strategy for the returned commits.

        Disables all sorting strategies, then enables only the single strategy supplied by the caller.

        Parameters:
        s - a sorting strategy to enable.
      • sort

        public void sort​(RevSort s,
                         boolean use)
        Add or remove a sorting strategy for the returned commits.

        Multiple strategies can be applied at once, in which case some strategies may take precedence over others. As an example, RevSort.TOPO must take precedence over RevSort.COMMIT_TIME_DESC, otherwise it cannot enforce its ordering.

        Parameters:
        s - a sorting strategy to enable or disable.
        use - true if this strategy should be used, false if it should be removed.
      • getRevFilter

        @NonNull
        public RevFilter getRevFilter()
        Get the currently configured commit filter.
        Returns:
        the current filter. Never null as a filter is always needed.
      • setRevFilter

        public void setRevFilter​(RevFilter newFilter)
        Set the commit filter for this walker.

        Multiple filters may be combined by constructing an arbitrary tree of AndRevFilter or OrRevFilter instances to describe the boolean expression required by the application. Custom filter implementations may also be constructed by applications.

        Note that filters are not thread-safe and may not be shared by concurrent RevWalk instances. Every RevWalk must be supplied its own unique filter, unless the filter implementation specifically states it is (and always will be) thread-safe. Callers may use RevFilter.clone() to create a unique filter tree for this RevWalk instance.

        Parameters:
        newFilter - the new filter. If null the special RevFilter.ALL filter will be used instead, as it matches every commit.
        See Also:
        AndRevFilter, OrRevFilter
      • getTreeFilter

        @NonNull
        public TreeFilter getTreeFilter()
        Get the tree filter used to simplify commits by modified paths.
        Returns:
        the current filter. Never null as a filter is always needed. If no filter is being applied TreeFilter.ALL is returned.
      • setTreeFilter

        public void setTreeFilter​(TreeFilter newFilter)
        Set the tree filter used to simplify commits by modified paths.

        If null or TreeFilter.ALL the path limiter is removed. Commits will not be simplified.

        If non-null and not TreeFilter.ALL then the tree filter will be installed. Commits will have their ancestry simplified to hide commits that do not contain tree entries matched by the filter, unless setRewriteParents(false) is called.

        Usually callers should be inserting a filter graph including TreeFilter.ANY_DIFF along with one or more PathFilter instances.

        Parameters:
        newFilter - new filter. If null the special TreeFilter.ALL filter will be used instead, as it matches everything.
        See Also:
        PathFilter
      • setRewriteParents

        public void setRewriteParents​(boolean rewrite)
        Set whether to rewrite parent pointers when filtering by modified paths.

        By default, when setTreeFilter(TreeFilter) is called with non- null and non-TreeFilter.ALL filter, commits will have their ancestry simplified and parents rewritten to hide commits that do not match the filter.

        This behavior can be bypassed by passing false to this method.

        Parameters:
        rewrite - whether to rewrite parents; defaults to true.
        Since:
        3.4
      • isRetainBody

        public boolean isRetainBody()
        Should the body of a commit or tag be retained after parsing its headers?

        Usually the body is always retained, but some application code might not care and would prefer to discard the body of a commit as early as possible, to reduce memory usage.

        True by default on RevWalk and false by default for ObjectWalk.

        Returns:
        true if the body should be retained; false it is discarded.
      • setRetainBody

        public void setRetainBody​(boolean retain)
        Set whether or not the body of a commit or tag is retained.

        If a body of a commit or tag is not retained, the application must call parseBody(RevObject) before the body can be safely accessed through the type specific access methods.

        True by default on RevWalk and false by default for ObjectWalk.

        Parameters:
        retain - true to retain bodies; false to discard them early.
      • isFirstParent

        public boolean isFirstParent()
        Returns:
        whether only first-parent links should be followed when walking.
        Since:
        5.5
      • setFirstParent

        public void setFirstParent​(boolean enable)
        Set whether or not only first parent links should be followed.

        If set, second- and higher-parent links are not traversed at all.

        This must be called prior to markStart(RevCommit).

        Parameters:
        enable - true to walk only first-parent links.
        Since:
        5.5
      • lookupBlob

        @NonNull
        public RevBlob lookupBlob​(AnyObjectId id)
        Locate a reference to a blob without loading it.

        The blob may or may not exist in the repository. It is impossible to tell from this method's return value.

        Parameters:
        id - name of the blob object.
        Returns:
        reference to the blob object. Never null.
      • lookupTree

        @NonNull
        public RevTree lookupTree​(AnyObjectId id)
        Locate a reference to a tree without loading it.

        The tree may or may not exist in the repository. It is impossible to tell from this method's return value.

        Parameters:
        id - name of the tree object.
        Returns:
        reference to the tree object. Never null.
      • lookupCommit

        @NonNull
        public RevCommit lookupCommit​(AnyObjectId id)
        Locate a reference to a commit without loading it.

        The commit may or may not exist in the repository. It is impossible to tell from this method's return value.

        See parseHeaders(RevObject) and parseBody(RevObject) for loading contents.

        Parameters:
        id - name of the commit object.
        Returns:
        reference to the commit object. Never null.
      • lookupTag

        @NonNull
        public RevTag lookupTag​(AnyObjectId id)
        Locate a reference to a tag without loading it.

        The tag may or may not exist in the repository. It is impossible to tell from this method's return value.

        Parameters:
        id - name of the tag object.
        Returns:
        reference to the tag object. Never null.
      • lookupAny

        @NonNull
        public RevObject lookupAny​(AnyObjectId id,
                                   int type)
        Locate a reference to any object without loading it.

        The object may or may not exist in the repository. It is impossible to tell from this method's return value.

        Parameters:
        id - name of the object.
        type - type of the object. Must be a valid Git object type.
        Returns:
        reference to the object. Never null.
      • lookupOrNull

        public RevObject lookupOrNull​(AnyObjectId id)
        Locate an object that was previously allocated in this walk.
        Parameters:
        id - name of the object.
        Returns:
        reference to the object if it has been previously located; otherwise null.
      • parseAny

        @NonNull
        public RevObject parseAny​(AnyObjectId id)
                           throws MissingObjectException,
                                  IOException
        Locate a reference to any object and immediately parse its headers.

        This method only returns successfully if the object exists and was parsed without error. Parsing an object can be expensive as the type must be determined. For blobs this may mean the blob content was unpacked unnecessarily, and thrown away.

        Parameters:
        id - name of the object.
        Returns:
        reference to the object. Never null.
        Throws:
        MissingObjectException - the supplied does not exist.
        IOException - a pack file or loose object could not be read.
      • parseAny

        public <T extends ObjectIdAsyncRevObjectQueue parseAny​(Iterable<T> objectIds,
                                                                 boolean reportMissing)
        Asynchronous object parsing.
        Parameters:
        objectIds - objects to open from the object store. The supplied collection must not be modified until the queue has finished.
        reportMissing - if true missing objects are reported by calling failure with a MissingObjectException. This may be more expensive for the implementation to guarantee. If false the implementation may choose to report MissingObjectException, or silently skip over the object with no warning.
        Returns:
        queue to read the objects from.
      • parseHeaders

        public void parseHeaders​(RevObject obj)
                          throws MissingObjectException,
                                 IOException
        Ensure the object's critical headers have been parsed.

        This method only returns successfully if the object exists and was parsed without error.

        Parameters:
        obj - the object the caller needs to be parsed.
        Throws:
        MissingObjectException - the supplied does not exist.
        IOException - a pack file or loose object could not be read.
      • parseBody

        public void parseBody​(RevObject obj)
                       throws MissingObjectException,
                              IOException
        Ensure the object's full body content is available.

        This method only returns successfully if the object exists and was parsed without error.

        Parameters:
        obj - the object the caller needs to be parsed.
        Throws:
        MissingObjectException - the supplied does not exist.
        IOException - a pack file or loose object could not be read.
      • peel

        public RevObject peel​(RevObject obj)
                       throws MissingObjectException,
                              IOException
        Peel back annotated tags until a non-tag object is found.
        Parameters:
        obj - the starting object.
        Returns:
        If obj is not an annotated tag, obj. Otherwise the first non-tag object that obj references. The returned object's headers have been parsed.
        Throws:
        MissingObjectException - a referenced object cannot be found.
        IOException - a pack file or loose object could not be read.
      • newFlag

        public RevFlag newFlag​(String name)
        Create a new flag for application use during walking.

        Applications are only assured to be able to create 24 unique flags on any given revision walker instance. Any flags beyond 24 are offered only if the implementation has extra free space within its internal storage.

        Parameters:
        name - description of the flag, primarily useful for debugging.
        Returns:
        newly constructed flag instance.
        Throws:
        IllegalArgumentException - too many flags have been reserved on this revision walker.
      • carry

        public void carry​(RevFlag flag)
        Automatically carry a flag from a child commit to its parents.

        A carried flag is copied from the child commit onto its parents when the child commit is popped from the lowest level of walk's internal graph.

        Parameters:
        flag - the flag to carry onto parents, if set on a descendant.
      • carry

        public void carry​(Collection<RevFlag> set)
        Automatically carry flags from a child commit to its parents.

        A carried flag is copied from the child commit onto its parents when the child commit is popped from the lowest level of walk's internal graph.

        Parameters:
        set - the flags to carry onto parents, if set on a descendant.
      • retainOnReset

        public final void retainOnReset​(RevFlag flag)
        Preserve a RevFlag during all reset methods.

        Calling retainOnReset(flag) avoids needing to pass the flag during each resetRetain() invocation on this instance.

        Clearing flags marked retainOnReset requires disposing of the flag with #disposeFlag(RevFlag) or disposing of the entire RevWalk by #dispose().

        Parameters:
        flag - the flag to retain during all resets.
        Since:
        3.6
      • retainOnReset

        public final void retainOnReset​(Collection<RevFlag> flags)
        Preserve a set of RevFlags during all reset methods.

        Calling retainOnReset(set) avoids needing to pass the flags during each resetRetain() invocation on this instance.

        Clearing flags marked retainOnReset requires disposing of the flag with #disposeFlag(RevFlag) or disposing of the entire RevWalk by #dispose().

        Parameters:
        flags - the flags to retain during all resets.
        Since:
        3.6
      • disposeFlag

        public void disposeFlag​(RevFlag flag)
        Allow a flag to be recycled for a different use.

        Recycled flags always come back as a different Java object instance when assigned again by newFlag(String).

        If the flag was previously being carried, the carrying request is removed. Disposing of a carried flag while a traversal is in progress has an undefined behavior.

        Parameters:
        flag - the to recycle.
      • reset

        public final void reset()
        Resets internal state and allows this instance to be used again.

        Unlike dispose() previously acquired RevObject (and RevCommit) instances are not invalidated. RevFlag instances are not invalidated, but are removed from all RevObjects.

      • resetRetain

        public final void resetRetain​(RevFlagSet retainFlags)
        Resets internal state and allows this instance to be used again.

        Unlike dispose() previously acquired RevObject (and RevCommit) instances are not invalidated. RevFlag instances are not invalidated, but are removed from all RevObjects.

        Parameters:
        retainFlags - application flags that should not be cleared from existing commit objects.
      • resetRetain

        public final void resetRetain​(RevFlag... retainFlags)
        Resets internal state and allows this instance to be used again.

        Unlike dispose() previously acquired RevObject (and RevCommit) instances are not invalidated. RevFlag instances are not invalidated, but are removed from all RevObjects.

        See retainOnReset(RevFlag) for an alternative that does not require passing the flags during each reset.

        Parameters:
        retainFlags - application flags that should not be cleared from existing commit objects.
      • reset

        protected void reset​(int retainFlags)
        Resets internal state and allows this instance to be used again.

        Unlike dispose() previously acquired RevObject (and RevCommit) instances are not invalidated. RevFlag instances are not invalidated, but are removed from all RevObjects. The value of firstParent is retained.

        Parameters:
        retainFlags - application flags that should not be cleared from existing commit objects.
      • dispose

        public void dispose()
        Dispose all internal state and invalidate all RevObject instances.

        All RevObject (and thus RevCommit, etc.) instances previously acquired from this RevWalk are invalidated by a dispose call. Applications must not retain or use RevObject instances obtained prior to the dispose call. All RevFlag instances are also invalidated, and must not be reused.

      • iterator

        public Iterator<RevCommit> iterator()

        Returns an Iterator over the commits of this walker.

        The returned iterator is only useful for one walk. If this RevWalk gets reset a new iterator must be obtained to walk over the new results.

        Applications must not use both the Iterator and the next() API at the same time. Pick one API and use that for the entire walk.

        If a checked exception is thrown during the walk (see next()) it is rethrown from the Iterator as a RevWalkException.

        Specified by:
        iterator in interface Iterable<RevCommit>
        See Also:
        RevWalkException
      • assertNotStarted

        protected void assertNotStarted()
        Throws an exception if we have started producing output.
      • assertNoCommitsMarkedStart

        protected void assertNoCommitsMarkedStart()
        Throws an exception if any commits have been marked as start.

        If markStart(RevCommit) has already been called, reset() can be called to satisfy this condition.

        Since:
        5.5
      • toObjectWalkWithSameObjects

        public ObjectWalk toObjectWalkWithSameObjects()
        Create and return an ObjectWalk using the same objects.

        Prior to using this method, the caller must reset this RevWalk to clean any flags that were used during the last traversal.

        The returned ObjectWalk uses the same ObjectReader, internal object pool, and free RevFlags. Once the ObjectWalk is created, this RevWalk should not be used anymore.

        Returns:
        a new walk, using the exact same object pool.
      • createCommit

        protected RevCommit createCommit​(AnyObjectId id)
        Construct a new unparsed commit for the given object.
        Parameters:
        id - the object this walker requires a commit reference for.
        Returns:
        a new unparsed reference for the object.
      • assumeShallow

        public void assumeShallow​(Collection<? extends ObjectId> ids)
        Assume additional commits are shallow (have no parents).

        This method is a No-op if the collection is empty.

        Parameters:
        ids - commits that should be treated as shallow commits, in addition to any commits already known to be shallow by the repository.
        Since:
        3.3