Files
scylladb/cdc
Piotr Dulikowski 534e9ba27d cdc: store information on ttl in "ttl" column, not in tuples
This patch changes the way TTL is stored in the CDC log table. Instead
of including TTL of cell `X` in the third element of the tuple in column
`_X`, TTL is written to the previously unused column `ttl`. This is done
for cosmetic purposes.

This implementation works under assumption that there will be only one
TTL included in a mutation coming from a CQL write. This might not be
the case when writing a batch that modifies the same row twice, e.g.:

```
BATCH
INSERT INTO ks.t (pk, ck, v1) VALUES (1,2,3) USING TTL 10;
INSERT INTO ks.t (pk, ck, v2) VALUES (1,2,3) USING TTL 20;
END BATCH
```

In this case, this implementation will choose only one TTL value to be
written in the CDC log:

```
... | batch_seq_no | _ck | _pk | _v1    | _v2    | operation | ttl
...-+--------------+-----+-----+--------+--------+-----------+-----
... |            0 |   2 |   1 | (0, 3) | (0, 3) |         1 |  20
```

This behavior might be changed as a part of issue #5719, which considers
splitting a batch write mutation when it contains multiple writes to the
same row.

Refs #5689
Tests: unit(dev)
2020-02-08 11:10:09 +02:00
..
2020-02-06 08:00:01 +01:00