v1.0.0

Improvements

• Allow doing Mutate and Alter Operations using Ratel(Dgraph browser).
• Export now uses correct blank node syntax.
• DropAttr now also removes the schema for the attribute (previously it just removed the edges).
• Tablet metadata is removed from Zero after deletion of predicate.
• LRU size is changed dynamically now based on max_memory_mb
• RunValueLogGC is called for every GB increase in size of value logs.
• Prohibit string to password schema change.
• Make purging less aggressive.
• Check if GraphQL Variable is defined before using.

Bugfixes

• Language tag parsing in queries now accepts digits (in line with RDF parsing).
• Ensure that GraphQL variables are declared before use.
• Membership stream doesn't get stuck if node steps down as leader.
• Fix issue where sets were not being returned after doing a S P * deletion when part of same
  transaction.
• Empty string values are stored as it is and no strings have special meaning now.
• Correctly update order of facetMatrix when orderdesc/orderasc is applied.
• Allow live and bulk loaders to work with multiple zeros.
• Fix sorting for predicates with multiple language tags.
• Fix alias edge cases in normalize directive.
• Allow reading new index key mutated as part of same transaction.
• Fix bug in value log GC in badger.
• SIGINT now forces a shutdown after 5 seconds when there are pending RPCs.

v0.9.4

This release contains a small fix which allows using the data directories from previous versions(v0.9.0-v0.9.2).
Users should skip v0.9.3 and upgrade to v0.9.4 instead.

Improvement

• Max size for GRPC messages that can be sent by the server is now set to 4GB.

v0.9.3

This release contains important bugfixes and improvements.

Features

• Support for alias while asking for facets.
• Support for general configuration via environment variables and configuration files.
• IgnoreIndexConflict field in api.Mutation which allows ignoring conflicts on index keys.

Improvements

• expand(_all_) now correctly gives all language variants of a string.
• When showing a predicate with list type, only values without a language tag are shown. To get the values of the predicate that are tagged with a language, query the predicate with that language explicitly.
• expand(_all_) now follows reverse edges.
• Don't return uid for nodes without any children when requested through debug flag.
• GraphQL variables for HTTP endpoints. Variable map can be set as a JSON
  object using the X-Dgraph-Vars header.
• Live loader treats subjects/predicates that look like UIDs as existing nodes
  rather than new nodes.
• Proto definitions are split into intern and api. Users using v0.9.x of the Go client will have to change their code to use api instead of proto as package name.

Bugfixes

• Indexes now correctly maintained when deleting via S * * and S P *.
• Validate the address advertised by dgraph nodes.
• Store/Restore peer map on snapshot.
• Fix rdfs per second reporting in live loader.
• Fix bug in @groupby queries where predicate was converted to lower case.
• Fix bug in lru eviction.

Dgraph v0.9.2 Release

This release contains bugfixes, features and minor improvements.

Features

• Support for removing dead nodes from the quorum.
• Support for alias in groupby queries.
• DeleteEdges helper function for Go client.

Improvements

• Dgraph tries to abort long running/abandoned transactions.
• Reduce dependencies for the Go client.
• depth and loop arguments are now passed inside @recurse.

Bugfixes

• base36 encode keys which are part of the TxnContext before sending to the client. This makes sure they are valid UTF-8 strings and the client can parse them.
• Fix --ui flag parsing.
• Fix TLS flag parsing for Dgraph server and live loader.
• Fix race condition in expand(_all_) queries. #1783

Dgraph v0.9.1 Release

This release contains a couple of small changes in the HTTP API.

• X-Dgraph-StartTs header goes away. The StartTs is now passed as a path parameter instead of a header.
• For /commit API, keys are passed in the body instead of X-Dgraph-Keys header.

The HTTP docs contain elaborate examples.

Dgraph v0.9.0 Release

The released binaries would not work with data from older versions. You can follow these instructions to upgrade Dgraph.

The latest release has a lot of breaking changes but also brings powerful features like Transactions, support for CJK and custom tokenization.

Note: Starting from this release, every read and write is done via transactions. You could run read-only transactions, read-write or write-only transactions; but there's no way to avoid transactional behavior.

Features

• Dgraph adds support for distributed ACID transactions (a blog post is in works). Transactions can be done via the Go, Java or HTTP clients (JS client coming). See docs here.
• Support for Indexing via Custom tokenizers.
• Support for CJK languages in the full-text index.

Breaking changes

Running Dgraph

• We have consolidated all the server, zero, live/bulk-loader binaries into a single dgraph binary for convenience. Instructions for running Dgraph can be found in the docs.
• For Dgraph server, Raft ids can be assigned automatically. A user can optionally still specify an ID, via --idx flag.
• --peer flag which was used to specify another Zero instance’s IP address is being replaced by --zero flag to indicate the address corresponds to Dgraph zero.
• port, grpc_port and worker_port flags have been removed from Dgraph server and Zero. The ports are:

• Internal Grpc: 7080
• HTTP: 8080
• External Grpc: 9080 (Dgraph server only)

Users can set port_offset flag, to modify these fixed ports.

Queries

• Queries, mutations and schema updates are done through separate endpoints. Queries can no longer have a mutation block.
• Queries can be done via Query Grpc endpoint (it was called Run before) or the /query HTTP handler.
• _uid_ is renamed to uid. So queries now need to request for uid. Example

{
  bladerunner(func: eq(name@en, "Blade Runner")) {
    uid
    name@en
  }
}

• Facets response structure has been modified and is a lot flatter. Facet key is now predicate|facet_name.
  Examples for Go client and HTTP.
• Query latency is now returned as numeric (ns) instead of string.
• Recurse is now a directive. So queries with recurse keyword at root won't work anymore.
• Syntax for count at root has changed. You need to ask for count(uid), instead of count().

Mutations

• Mutations can only be done via Mutate Grpc endpoint or via /mutate HTTP handler.
• Mutate Grpc endpoint can be used to set/ delete JSON, or set/ delete a list of NQuads and set/ delete raw RDF strings.
• Mutation blocks don't require the mutation keyword anymore. Here is an example of the new syntax.

{
  set {
    <name> <is> <something> .
    <hometown> <is> "San Francisco" .
  }
}

• Upsert directive and mutation variables go away. Both these functionalities can now easily be achieved via transactions.

Schema

• <*> <pred> <*> operations, that is deleting a predicate can't be done via mutations anymore. They need to be done via Alter Grpc endpoint or via the /alter HTTP handler.
• Drop all is now done via Alter.
• Schema updates are now done via Alter Grpc endpoint or via /alter HTTP handler.

Go client

• Query Grpc endpoint returns response in JSON under Json field instead of protocol buffer. client.Unmarshal method also goes away from the Go client. Users can use json.Unmarshal for unmarshalling the response.
• Response for predicate of type geo can be unmarshalled into a struct. Example here.
• Node and Edge structs go away along with the SetValue... methods. We recommend using SetJson and DeleteJson fields to do mutations.
• Examples of how to use transactions using the client can be found at https://docs.dgraph.io/clients/#go.

Embedded Dgraph

Embedded dgraph goes away. We haven’t seen much usage of this feature. And it adds unnecessary maintenance overhead to the code.

Others

• Dgraph live no longer stores external ids. And hence the xid flag is gone.

Bugfixes

• #1620
• #1646
• #1651
• #1659
• #1662
• #1758

Dgraph v0.8.3 Release

The released binaries would only work with data from v0.8.2. If you are upgrading from an older version, you can follow these instructions to upgrade Dgraph.

Features

• dgraphzero binary for controlling the cluster and moving data between nodes based on load.
• Allow doing mutations using SetObject and DeleteObject methods from the Go client.
• Add ability to delete all data from the database.

Improvements

• dgraph-bulk-loader(bulkloader) - Lot of memory improvements. The dgraph-bulk-loader is now faster than ever.
• Aggregate uids stored in a variable over levels for recurse query.
• Return root key in response even if there are no results. #1533
• Enable expand(_all_) with recurse.
• Remove ObjectType from NQuad proto message.
• Offical support for precompiled Windows binaries.

Bugfixes

• #1565 Fix bugs in gql/parser thanks to @pborzenkov.
• #1558 Prevent nil pointer crash thanks to @noh4ck
• #1524
• #1517
• #1526
• #1564
• #1577
• #1571
• #1597
• #1594
• #1598
• #1587

Dgraph v0.8.2 Release

The released binaries wouldn't work with data from older versions. You can follow these instructions to upgrade Dgraph from an older version.

Features

• Bulkloader - Load data into Dgraph at least 5x faster.
• Allow setting multiple scalar values.
• Support for MultiPolygon type for Geo values.
• Add ability to sort by multiple predicates.
• Allow ability to upsert nodes using eq function.

Improvements

• Backup requests for intracluster retrievals. If the first one doesn't finish in 10ms, we try another server.
• Make reads/writes follow the requirements of linearizability.
• Allow value variables in inequality functions.

Bugfixes

• . language handling in functions. #1337
• Recurse query panics. #1350
• Variable not populated or missing. #1356
• Panic in appendDummyValues. #1359
• Write hard-state after persisting entries. #1365
• Can't query schema using gRPC interface. #1369
• Facet matrix not in sync with uid matrix after filtering. #1362
• Empty predicate crashes db. #1381
• Mutation with object vars crashes DB. #1382
• Sort and validate facets coming from clients. #1419
• Check for empty predicate in schema mutation. #1447
• Filter may override root query. #1455
• #1444
• #1427
• #1474
• #1464
• #1477

Dgraph v0.8.1 Release

The released binaries wouldn't work with data from older versions. You can follow these instructions to upgrade Dgraph from an older version.

Changes from v0.8.0

• It's mandatory to specify the type of the tokenizer while specifying an index in the schema.
• Renamed dateTime tokenizer to year.
• It's mandatory to specify the memory_mb flag while starting Dgraph.
• JSON response is nested under data and errors key in accordance with the GraphQL spec.

Features

• Ability to sort by facets.
• Support for checkpointing in Dgraphloader.
• Allow aggregations at the top level of the query.

Improvements

Server

• Optimizations in JSON encoding so that it uses lesser memory (157x) and is faster (512x).
• Optimize Inequality filters. Fetch data keys directly and compare if their count is less than the number of index keys to be fetched.
• Return JSON results in the same order as the query.
• Get rid of sync.Pool in postings which stabilizes memory used by the Dgraph process.
• Compress posting lists using bit packing and delta encoding with simd instructions so that it uses lesser memory.

Client

• Helper function to delete a predicate.
• Helper function to delete a node.
• Helper function to delete an edge.
• Allow setting string values with lang tags.
• Make BatchMutation.Flush() retry logic tunable for Go client through MaxRetries. (#796)

Bugfixes

• Return count even if it's zero. (#1212)
• uid_in function should accept hexidecimal for uid. (#1204)
• Dgraphloader should allow spaces between comma separated files. (#1209)
• Fix bug when using variables in mutations. (#1217)
• Disallow repeated arguments. (#1223)
• Fix bad memory usage while serializing to JSON. (#982, #1138, #1227)
• Ensure variables for sorting is a value variable. (#1235)
• Clear out struct given in client.Unmarshal before unmarshalling response using the Go client. (#1216)
• Fix Maximum call stack size exceeded error on UI. (#1201)
• Delete reverse edge when doing S P * deletion. (#1271)
• Panic runtime error. (#1284)
• Count index not updated after schema update. (#1247)
• Slow start in embedded mode. (#1246)
• Throw an error if a variable is used before definition. (#1322)
• Throw an error if NQuads are not separated by a newline. (#1320)
• Limit number of results returned as part of @normalize. (#1268)
• Double click on web UI cyclic graph causes an infinite loop. (#1302)
• Incorrect function behaviour when language is not specified. (#1295)
• Stop storing responses in Dgraph browser. Fix localStorage quota reached error. (#981)

Dgraph v0.8.0 Release

This release would not work on v0.7 data directly. Needs backup and reimport.

Features

Server

• Suppport has function.
• Support for variables in mutation blocks.
• Support for k shortest paths.
• Support deletion of predicate.
• Support for uid_in function.
• ignoreReflex directive to remove reflexive edges from results.
• Allow to use Dgraph as a library.
• Ability to monitor server metrics.

Client

• Go client now supports an Unmarshal function which allows unmarshalling the protocol buffer response to a custom struct.
• Add support for storing and retrieving raw bytes.
• Cleaner and more intuitive API for doing mutations.

Improvements

• Move away from RocksDB and get rid of CGO. Start using Badger.
• Assign uids sequentially.
• Allow retrieving other children with @groupby.
• Allow aliases for math expressions.
• Support filtering using multiple tokens with eq.
• Better language lists semantics. Issue #1010
• Add a uid function which can be used for filtering uids at any level.
• Allow adding @count indexes which allow for faster count comparison queries at the root.
• Optimisations in how we store Posting List data structure. Store uids separately.
• Allow specifying a max depth argument for the shortest path.
• Support for month, day, hour indexes for dateTime datatype.
• Get rid of lhmap and add an LRU Cache. We don't do Stop the World anymore.

Changes (from v0.7.7)

• date type is no longer supported as a schema type in Dgraph. You should use dateTime instead.
• Change of syntax for querying by id.

# Before
{
  me(id: [0x1, 0x2]) {
    ...
  }
}

# Now
{
  me(func: uid(0x1, 0x2)) {
    ...
  }
}

• Dgraph doesn't fingerprint xids anymore. They are now treated similarly to other edges. So now if you had an xid and you wanted to query by it you would use.

# You need to have an index on xid for calling `eq` function.
{
  me(func: eq(xid, "alice")) {
    ....
  } 
}

• /admin/backup is now /admin/export. It outputs blank nodes which can now be used to load data into a fresh instance.
• There is a separate port for http (default 8080) and grpc (default 9080) now.

Bugfixes

• Remove concurrent filtering data race. Issue #1020
• Fix panic issues related to tracing. Wait for all goroutines to finish before retuning error.
• Fix crash caused by nil list. #1035
• #793
• #939
• #960
• #993
• #1007
• #1025
• #1118
• #1157
• #1163