Hello fellow devs,
I was out the past couple Fridays, so that is why this will be a big
reporting period :)
There will not be a report next week since it's the day after
Thanksgiving in the US.

In case you missed it, AsterixDB 1.0.0 RC0 is out. Please give it a
try- there is a lot of new functionality, and 2 new packaging methods
(Docker and Helm) that need scrutiny. I have tried them myself, but
"works on my machine" isn't a good metric.

Additionally, the patch which enables GHA was merged and seems to be
working well on master. This means Gerrit and Jenkins will soon be
deprecated. I will send a separate email with further details on the
exact timing, however, there will be a day when no new patches can be
submitted to Gerrit soon. All changes must go through GitHub PRs. This
is because Gerrit kind of expects that it is the source of truth, and
merging changes outside of it tends to cause issues (like a
netsplit/split brain).

Looking forward to hearing about what everyone has been upto since Halloween!

Toplevel status:
- 1.0.0 RC0 Out!
- Switch over to PRs and Actions impending

Issue roundup:
New:
- 3981 (Columns are being fetched for the mega leaf being rejected)
- 3982 (UPDATE can ignore specified dataverse)
- 3983 (Pushdowns can use wrong schema for array idx and computed fields)
- 3984 (Channel alloc scans every slot of a multiplex conn)
- 3985 (External Functions)
- 3986 (MuxDemux registers accepted conns without holding lock)
- 3987 (Pushdown can narrow a path consumed w/o var)
- 3988 (Window aggregates duplicate a shared path in the nested plan)
- 3989 (Type checks give wrong answers on a value narrowed by field pushdown)
- 3990 (FABN misses closed fields when the name is computed)
- 3991 (Score k-means|| rounds against new candidates only)
- 3992 (Bound columnar schemas for objects whose field names are data)
- 3993 (Add approximate aggregate functions)
- 3994 (Index used through OBJECT_* calls that alter the indexed field)
- 3995 (OBJECT_ADD writes an existing open field into a closed slot)
- 3996 (OBJECT_REMOVE_FIELDS fails to compile with a computed field name)
- 3997 (Function resources SerDe is incorrect)
- 3998 (S3 parallel dl rejects clients the serial one accepts)
- 3999 (Cache B-Tree indexes page-wise in cloud mode with lazy caching)
- 4000 (A number as a WITH clause fails to compile)
- 4001 (Shredded VARIANT object reads as null when only untyped fields
are req'd)
- 4002 (Variant projection pushdown can clip the wrong col)
- 4003 (Timestamps in Iceberg can be rounded to before unix epoch)

Fixed/Implemented:
- 3981 (Columns are being fetched for the mega leaf being rejected)
- 3982 (UPDATE can ignore specified dataverse)
- 3984 (Channel alloc scans every slot of a multiplex conn)
- 3986 (MuxDemux registers accepted conns without holding lock)
- 3991 (Score k-means|| rounds against new candidates only)
- 3992 (Bound columnar schemas for objects whose field names are data)
- 3997 (Function resources SerDe is incorrect)
- 3998 (S3 parallel dl rejects clients the serial one accepts)

Open APEs:
Schema inference: Patch rebased and comments addressed

Codegen:
No activity in a while

Sampling cursor:
APE exists in cwiki but needs vote thread on dev@

External functions:
Amendments accepted, work in progress. Needs vote.

Dynamic memory
WIP Patch forthcoming?

MCP
Need input from Suryaa

Columnar schemas for unbounded object keys (APE 37)
Needs vote thread

Blocked-ish patches:
- Geometry type for Columnar (after Nov?)
- Docs improvements (still! but hopefully for 0.9.10/1.0.0)
- Metadata function introspection (after Nov?)

Reply via email to