Steve McCanne

The ZSON serializer has a bug where it thinks a named type is implied when it isn't. Reproduce with this input: ``` $ echo '[1(=A),{s:"foo"(=B)}(=A)]' | zq - [1(=A),{s:"foo"}(=A)] ```...

parallelize streams

1

For queries that are flow insensitive, the same parallelization logic that we use for lake can also apply to streams (e.g., zq or internal readers like ingest). We should logic...

furl/unfurl

1

furl/unfurl shall be introduced as functions in Zed and return an array of records comprising a single path element (as in flatten) called `key` and the corresponding value derived from...

A function call path shall take an l-value expression and return the encoded path as an array of elements. e.g., ``` path(a.b[2].c) ``` returns ``` ["a", "b", 2, "c"] ```...

`flatten()` currently has inconsistent behavior between record paths and array paths. In this issue, we introduce the concept of a jq-like generalized path where a path defines any route through...

replace <tag> with <name>

1

Noah made this suggestion: ``` We might want to follow Git's lead here and use "name" in place of "tag" for these terms, for two reasons. First, a Git user's...

handle multiple pool keys

5

e.g., primary, second etc keys should be allowed for pools. The data structures are all set up for this but the compiler and optimizer currently assume a single primary key....

This issue builds on the vector's filter + projection capability with aggregations and group-bys. Here filtering and selection of fields will create vectors that are then fed into a native...

This issue builds on #4818 by adding filtering to the vector engine's projection capability. Here simple filtering of fields (with boolean logic) can be pushed into the vector scanner where...

The first step toward a working vector engine is simple projection. In this issue, we will wire the vector engine into the runtime so that it is actually invocable from...