Cherry-pick changes required for 1.2.2 release. #4951
Conversation
Special predicate names (i.e using Chinese characters) need to be wrapped around "<" and ">" brackets but the lexer does not currently support that. This PR adds support for that syntax. Fixes #4933
When the warning to delete and re-add the value was removed, non-list uid predicates were not being overwritten. This fixes this by iterating over the list and replacing theexisting postings with a copy with the operation set to a delete. It also adds a test to prevent this from happening again. Fixes #4879
String facet values are converted to a date time value if they match a time format. The documentation should mention this and other type conversions.
Currently, the host is only checked when the default credentials are used. This means using custom credentials will not work if there's no host in the destination URI as the logic that handles this case is not executed. This PR changes the logic so that these checks are always executed, regardless of the status of the credentials.
If the list is too big and is not split before writing it too disk, we could end up in the situation described in #4733. So the posting lists should be split before writing them to disk.
A language list with the value "*:." is lexed correctly and an error saying that if all the languages (the meaning of the character *) are requested, the language list should only have one element. But a language list with the value ".:*" fails with a lexing error. This change changes the lexer so that the two lists fail with the same error.
< > brackets are needed around the type name. They were missing.
martinmr
requested review from
danielmai,
manishrjain,
MichaelJCompton,
pawanrawal and
a team
as code owners
Trying to skip over the split keys sometimes skips over keys belonging to a different split key. ReadPostingList was changed to return a nil list when trying to read a part of a split key. All the operations on a nil list are now a no-op. This is a fix just for this release as the actual fix requires changes to the data format.
martinmr
force-pushed
the
martinmr/cp-changes-1.2.2
branch
from
710d4db to
1e6334e
Compare
There was a problem hiding this comment.
Reviewed 13 of 16 files at r1, 1 of 3 files at r2, 2 of 2 files at r3.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @danielmai, @martinmr, @MichaelJCompton, and @pawanrawal)
posting/list.go, line 777 at r3 (raw file):
// As the list grows, existing parts might be split if they become too big. func (l *List) Rollup() ([]*bpb.KV, error) { if l == nil {
Is this required? In our discussions, we decided against it.
worker/export.go, line 337 at r3 (raw file):
x.Check2(builder.WriteString("\t")) // While exporting type definitions, "<" and ">" brackets must be written around // the name of everse predicates or Dgraph won't be able to parse the exported schema.
Why does reverse predicate show up in exported data?
Also, there's a typo here -- reverse is missing an r.
There was a problem hiding this comment.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @danielmai, @golangcibot, @manishrjain, @MichaelJCompton, and @pawanrawal)
posting/list.go, line 39 at r2 (raw file):
Previously, golangcibot (Bot from GolangCI) wrote…
File is not
gofmt-ed with-s(fromgofmt)"github.com/golang/protobuf/proto" "github.com/pkg/errors"
Done.
posting/list.go, line 777 at r3 (raw file):
Previously, manishrjain (Manish R Jain) wrote…
Is this required? In our discussions, we decided against it.
The one in rollup is. ReadPostingList returns a nil list so it's important that Rollup works with a nil list. This will cause the split keys to be skipped.
I did the rest of the public methods just to be safe.
worker/export.go, line 337 at r3 (raw file):
Previously, manishrjain (Manish R Jain) wrote…
Why does reverse predicate show up in exported data?
Also, there's a typo here -- reverse is missing an r.
Types can have fields that are reverse predicates.
I will not squash this PR so the log is preserved.
My changes to the bulk loader are missing because the bulk loader has a lot of changes
and not all of them have been cherry-picked into this branch. Not sure what to do about
that since I didn't make all of those changes so I am not sure exactly what commits should
be cherry-picked.
This change is
Docs Preview: