Reduces coupling. User's should not rely on the fact that it's an
std::map<>. It also allows us to extend row's interface with
domain-specific methods, which are a lot easier to discover than free
functions.
If method doesn't want to share schema ownership it doesn't have to
take it by shared pointer. The benefit is that it's slightly cheaper
and those methods may now be called from places which don't own
schema.
Deleted cells store deletion time not expiry time. This change makes
expiry() valid only for live cells with TTL and adds deletion_time(),
which is inteded to be used with deleted cells.
The immediate motivation for introducing frozen_mutation is inability
to deserialize current "mutation" object, which needs schema reference
at the time it's constructed. It needs schema to initialize its
internal maps with proper key comparators, which depend on schema.
frozen_mutation is an immutable, compact form of a mutation. It
doesn't use complex in-memory strucutres, data is stored in a linear
buffer. In case of frozen_mutation schema needs to be supplied only at
the time mutation partition is visited. Therefore it can be trivially
deserialized without schema.
In preparation for multiple memtables, move column_family::partitions into
its own class, and forward relevant calls from column_family.
A testonly_all_memtables() function was added to support sstable_test.
Currently we use the first byte of the token for determining the local
shard. This is suboptimal for two reasons:
1. the first bytes of the token were already used to select the node,
so they are not randomly distributed
2. using a single byte is not sufficient for large core counts, as the
modulo operation will not return evenly distributed results
Fix by using the final two bytes of the token.
A lookup can cause several data sources to be merged, in which case we will
have to return a temporary (containing data from all the data sources).
For simplicity, we start by always returning a temporary.
Ensure that read-side accessors are const. This is important in preparation
for multiple memtables (and later, sstables) since a read-side
mutation_partition may be a temporary object coming from multiple memtables
(and sstables) while a write-side mutation_partition is guaranteed to belong
to a single memtable (and thus, not be temporary).
Since writers will want non-const mutation_partitions to write to, they won't
be able to use the read-side accessors by accident.
Some tests (eg murmur_hash_test) need only byte manipulation
functions. By specifying dependencies precisely we can drastically
reduce recompilation times, which speeds up development cycle.
I managed to reduce recompilation time for murmur_hash_test from 5
minutes to 4 seconds by breaking dependency on whole urchin object
set.
Use commit log in database, from Calle:
"Initial" usage of the commitlog in database mutation path.
A commitlog is created in "work" dirs when initing the db
from a datadir. However, since we have neither disk data storage,
nor replay capability yet (and no real db config), the settings
are basically to just write in-memory serialization, write them to
disk and then discard them. So in fact, pointless. But at least using
the log...
* A commitlog is created in "work" dirs when initing the db
from a datadir. However, since we have neither disk data storage,
nor replay capability yet (and no real db config), the settings
are basically to just write in-memory serialization, write them to
disk and then discard them. So in fact, pointless. But at least using
the log...
* Moved the actual "apply" of mutation into database. If a commitlog
is active, add an entry to it before applying mutation.
Partitions should be ordered using Origin's ordering, which is first
by token, then by Origin's representation of the key. That is the
natural ordering of decorated_key.
This also changes mutation class to hold decorated_key, to avoid
decoration overhead at different layers.
This gives about 30% increase in tps in:
build/release/tests/perf/perf_simple_query -c1 --query-single-key
This patch switches query result format from a structured one to a
serialized one. The problems with structured format are:
- high level of indirection (vector of vectors of vectors of blobs), which
is not CPU cache friendly
- high allocation rate due to fine-grained object structure
On replica side, the query results are probably going to be serialized
in the transport layer anyway, so this change only subtracts
work. There is no processing of the query results on replica other
than concatenation in case of range queries. If query results are
collected in serialized form from different cores, we can concatenate
them without copying by simply appending the fragments into the
packet. This optimization is not implemented yet.
On coordinator side, the query results would have to be parsed from
the transport layer buffers anyway, so this also doesn't add work, but
again saves allocations and copying. The CQL server doesn't need
complex data structures to process the results, it just goes over it
linearly consuming it. This patch provides views, iterators and
visitors for consuming query results in serialized form. Currently the
iterators assume that the buffer is contiguous but we could easily
relax this in future so that we can avoid linearization of data
received from seastar sockets.
The coordinator side could be optimized even further for CQL queries
which do not need processing (eg. select * from cf where ...) we
could make the replica send the query results in the format which is
expected by the CQL binary protocol client. So in the typical case the
coordinator would just pass the data using zero-copy to the client,
prepending a header.
We do need structure for prefetched rows (needed by list
manipulations), and this change adds query result post-processing
which converts serialized query result into a structured one, tailored
particularly for prefetched rows needs.
This change also introduces partition_slice options. In some queries
(maybe even in typical ones), we don't need to send partition or
clustering keys back to the client, because they are already specified
in the query request, and not queried for. The query results hold now
keys as optional elements. Also, meta-data like cell timestamp and
ttl is now also optional. It is only needed if the query has
writetime() or ttl() functions in it, which it typically won't have.
