github.com/dgraph-io/dgraph

Changed

• Breaking changes for expand() queries
  • Remove expand(_forward_) and expand(_reverse_). (#4119)
  • Change expand(_all_) functionality to only include the predicates in the type. (#4171)
• Add support for Go Modules. (#4146)
• Simplify type definitions: type definitions no longer require the type (string, int, etc.) per field name. (#4017)
• Adding log lines to help troubleshoot snapshot and rollup. (#3889)
• Add --http flag to configure pprof endpoint for live loader. (#3846)
• Use snappy compression for internal gRPC communication. (#3368)
• Periodically run GC in all dgraph commands. (#4032, #4075)
• Exit early if data files given to bulk loader are empty. (#4253)
• Add support for first and offset directive in has function. (#3970)
• Pad encData to 17 bytes before decoding. (#4066)
• Remove usage of deprecated methods. (#4076)
• Show line and column numbers for errors in HTTP API responses. (#4012)
• Do not store non-pointer values in sync.Pool. (#4089)
• Verify that all the fields in a type exist in the schema. (#4114)
• Update badger to version v2.0.0. (#4200)
• Introduce StreamDone in bulk loader. (#4297)

Enterprise features:

• ACL: Disallow schema queries when an user has not logged in. (#4107)
• Block delete if predicate permission is zero. Fixes #4265. (#4349)

Added

• Support @cascade directive at subqueries. (#4006)
• Support @normalize directive for subqueries. (#4042)
• Support val() function inside upsert mutations (both RDF and JSON). (#3877, #3947)
• Support GraphQL Variables for facet values in @facets filters. (#4061)
• Support filtering by facets on values. (#4217)
• Add ability to query expand(TypeName) only on certain types. (#3920)
• Expose numUids metrics per query to estimate query cost. (#4033)
• Upsert queries now return query results in the upsert response. (#4269, #4375)
• Add support for multiple mutations blocks in upsert blocks. (#4210)
• Add total time taken to process a query in result under "total_ns" field. (#4312)

Enterprise features:

• Add encryption-at-rest. (#4351)

Removed

• Breaking change: Remove @type directive from query language. To filter an edge by a type, use @filter(type(TypeName)) instead of @type(TypeName). (#4016)

Enterprise features:

• Remove regexp ACL rules. (#4360)

Fixed

• Avoid changing order if multiple versions of the same edge is found.
• Consider reverse count index keys for conflict detection in transactions. Fixes #3893. (#3932)
• Clear the unused variable tlsCfg. (#3937)
• Do not require the last type declaration to have a new line. (#3926)
• Verify type definitions do not have duplicate fields. Fixes #3924. (#3925)
• Fix bug in bulk loader when store_xids is true. Fixes #3922. (#3950)
• Call cancel function only if err is not nil. Fixes #3966. (#3990)
• Change the mapper output directory from $TMP/shards to $TMP/map_output. Fixes #3959. (#3960)
• Return error if keywords used as alias in groupby. (#3725)
• Fix bug where language strings are not filtered when using custom tokenizer. Fixes #3991. (#3992)
• Support named queries without query variables. Fixes #3994. (#4028)
• Correctly set up client connection in x package. (#4036)
• Fix data race in regular expression processing. Fixes #4030. (#4065)
• Check for n.Raft() to be nil, Fixes #4053. (#4084)
• Fix file and directory permissions for bulk loader. (#4088)
• Ensure that clients can send OpenCensus spans over to the server. (#4144)
• Change lexer to allow unicode escape sequences. Fixes #4157.(#4175)
• Handle the count(uid) subgraph correctly. Fixes #4038. (#4122)
• Don't traverse immutable layer while calling iterate if deleteBelowTs > 0. Fixes #4182. (#4204)
• Bulk loader allocates reserved predicates in first reduce shard. Fixes #3968. (#4202)
• Only allow one alias per predicate. (#4236)
• Change member removal logic to remove members only once. (#4254)
• Disallow uid as a predicate name. (#4219)
• Drain apply channel when a snapshot is received. (#4273)
• Added RegExp filter to func name. Fixes #3268. (#4230)
• Acquire read lock instead of exclusive lock for langBaseCache. (#4279)
• Added proper handling of int and float for math op. #4132. (#4257)
• Don't delete group if there is no member in the group. (#4274)
• Sort alphabets of languages for non indexed fields. Fixes [#4005][]. (#4260)
• Copy xid string to reduce memory usage in bulk loader. (#4287)
• Adding more details for mutation error messages with scalar/uid type mismatch. (#4317)
• Limit UIDs per variable in upsert. Fixes #4021. (#4268)
• Return error instead of panic when geo data is corrupted. Fixes #3740. (#4318)
• Use txn writer to write schema postings. (#4296)
• Fix connection log message in dgraph alpha from "CONNECTED" to "CONNECTING" when establishing a connection to a peer. Fixes #4298. (#4303)
• Fix segmentation fault in backup. (#4314)
• Close store after stoping worker. (#4356)
• Don't pre allocate mutation map. (#4343)
• Cmd: fix config file from env variable issue in subcommands. Fixes #4311. (#4344)
• Fix segmentation fault in Alpha. Fixes #4288. (#4394)
• Fix handling of depth parameter for shortest path query for numpaths=1 case. Fixes #4169. (#4347)
• Do not return dgo.ErrAborted when client calls txn.Discard(). (#4389)
• Fix has pagination when predicate is queried with @lang. Fixes #4282. (#4331)
• Make uid function work with value variables in upsert blocks. Fixes #4424. (#4425)

Enterprise features:

• Fix bug when overriding credentials in backup request. Fixes #4044. (#4047)
• Create restore directory when running "dgraph restore". Fixes #4315. (#4352)
• Write group_id files to postings directories during restore. (#4365)