0
0
Fork 0
mirror of https://github.com/matrix-construct/construct synced 2024-11-05 05:18:54 +01:00
construct/matrix/events.cc

698 lines
12 KiB
C++
Raw Normal View History

// Matrix Construct
//
// Copyright (C) Matrix Construct Developers, Authors & Contributors
// Copyright (C) 2016-2019 Jason Volk <jason@zemos.net>
//
// Permission to use, copy, modify, and/or distribute this software for any
// purpose with or without fee is hereby granted, provided that the above
// copyright notice and this permission notice is present in all copies. The
// full license for this software is available in the LICENSE file.
namespace ircd::m::events
{
extern conf::item<size_t> dump_buffer_size;
}
decltype(ircd::m::events::dump_buffer_size)
ircd::m::events::dump_buffer_size
{
{ "name", "ircd.m.events.dump.buffer_size" },
{ "default", int64_t(512_KiB) },
};
void
ircd::m::events::rebuild()
{
static const event::fetch::opts fopts
{
event::keys::include {"type", "sender"}
};
static const m::events::range range
{
0, -1UL, &fopts
};
db::txn txn
{
*m::dbs::events
};
dbs::write_opts wopts;
wopts.appendix.reset();
wopts.appendix.set(dbs::appendix::EVENT_TYPE);
wopts.appendix.set(dbs::appendix::EVENT_SENDER);
size_t ret(0);
for_each(range, [&txn, &wopts, &ret]
(const event::idx &event_idx, const m::event &event)
{
wopts.event_idx = event_idx;
dbs::write(txn, event, wopts);
++ret;
if(ret % 8192UL == 0UL)
log::info
{
log, "Events type/sender table rebuild events %zu of %zu num:%zu txn:%zu %s",
event_idx,
vm::sequence::retired,
ret,
txn.size(),
pretty(iec(txn.bytes())),
};
return true;
});
log::info
{
log, "Events type/sender table rebuild events:%zu txn:%zu %s commit...",
ret,
txn.size(),
pretty(iec(txn.bytes())),
};
txn();
log::notice
{
log, "Events type/sender table rebuild complete.",
ret,
txn.size(),
pretty(iec(txn.bytes())),
};
}
void
ircd::m::events::dump__file(const string_view &filename)
{
fs::fd::opts fileopts(std::ios::out | std::ios::app);
fileopts.dontneed = true;
const fs::fd file
{
filename, fileopts
};
const unique_buffer<mutable_buffer> buf
{
size_t(dump_buffer_size)
};
char *pos{data(buf)};
size_t foff{0}, ecount{0}, acount{0}, errcount{0};
static const db::gopts gopts
{
db::get::NO_CACHE, db::get::NO_CHECKSUM
};
for(auto it(dbs::event_json.begin(gopts)); it; ++it) try
{
const event::idx seq
{
byte_view<uint64_t>(it->first)
};
const string_view source
{
it->second
};
const auto remain
{
size_t(data(buf) + size(buf) - pos)
};
assert(remain >= event::MAX_SIZE && remain <= size(buf));
const mutable_buffer mb
{
pos, remain
};
pos += copy(mb, source);
++ecount;
if(pos + event::MAX_SIZE > data(buf) + size(buf))
{
const const_buffer cb
{
data(buf), pos
};
foff += size(fs::append(file, cb));
pos = data(buf);
if(acount++ % 256 == 0)
{
char pbuf[48];
log::info
{
"dump[%s] %0.2lf%% @ seq %zu of %zu; %zu events; %s in %zu writes; %zu errors",
filename,
(seq / double(m::vm::sequence::retired)) * 100.0,
seq,
m::vm::sequence::retired,
ecount,
pretty(pbuf, iec(foff)),
acount,
errcount
};
}
}
}
catch(const ctx::interrupted &)
{
throw;
}
catch(const std::exception &e)
{
++errcount;
log::error
{
"dump[%s] events; %s in %zu writes; %zu errors :%s",
filename,
ecount,
acount,
errcount,
e.what(),
};
}
if(pos > data(buf))
{
const const_buffer cb{data(buf), pos};
foff += size(fs::append(file, cb));
++acount;
}
log::notice
{
log, "dump[%s] complete events:%zu using %s in writes:%zu errors:%zu",
filename,
ecount,
pretty(iec(foff)),
acount,
errcount,
};
}
bool
ircd::m::events::for_each(const range &range,
const event_filter &filter,
const closure &closure)
{
auto limit
{
json::get<"limit"_>(filter)?: 32L
};
return for_each(range, [&filter, &closure, &limit]
(const event::idx &event_idx, const m::event &event)
-> bool
{
if(!match(filter, event))
return true;
if(!closure(event_idx, event))
return false;
return --limit;
});
}
bool
ircd::m::events::for_each(const range &range,
const closure &closure)
{
event::fetch event
{
range.fopts? *range.fopts : event::fetch::default_opts
};
const bool ascending
{
range.first <= range.second
};
auto start
{
ascending?
range.first:
std::min(range.first, vm::sequence::retired)
};
const auto stop
{
ascending?
std::min(range.second, vm::sequence::retired + 1):
range.second
};
for(; start != stop; ascending? ++start : --start)
if(seek(std::nothrow, event, start))
if(!closure(start, event))
return false;
return true;
}
///TODO: This impl is temp. Need better dispatching based on filter before
///TODO: fetching event.
bool
ircd::m::events::for_each(const range &range,
const event_filter &filter,
const event::closure_idx_bool &closure)
{
auto limit
{
json::get<"limit"_>(filter)?: 32L
};
return for_each(range, event::closure_idx_bool{[&filter, &closure, &limit, &range]
(const event::idx &event_idx)
-> bool
{
const m::event::fetch event
{
std::nothrow, event_idx, range.fopts? *range.fopts : event::fetch::default_opts
};
if(!event.valid)
return true;
if(!match(filter, event))
return true;
if(!closure(event_idx))
return false;
return --limit;
}});
return true;
}
bool
ircd::m::events::for_each(const range &range,
const event::closure_idx_bool &closure)
{
const bool ascending
{
range.first <= range.second
};
auto start
{
ascending?
range.first:
std::min(range.first, vm::sequence::retired)
};
const auto stop
{
ascending?
std::min(range.second, vm::sequence::retired + 1):
range.second
};
auto &column
{
dbs::event_json
};
auto it
{
column.lower_bound(byte_view<string_view>(start))
};
for(; bool(it); ascending? ++it : --it)
{
const event::idx event_idx
{
byte_view<event::idx>(it->first)
};
if(ascending && event_idx >= stop)
break;
if(!ascending && event_idx <= stop)
break;
if(!closure(event_idx))
return false;
}
return true;
}
//
// events::state
//
bool
ircd::m::events::state::for_each(const closure &closure)
{
static const tuple none
{
{}, {}, {}, -1L, 0UL
};
return state::for_each(none, [&closure]
(const tuple &key) -> bool
{
return closure(key);
});
}
bool
ircd::m::events::state::for_each(const tuple &query,
const closure &closure)
{
db::column &column
{
dbs::event_state
};
char buf[dbs::EVENT_STATE_KEY_MAX_SIZE];
const string_view query_key
{
dbs::event_state_key(buf, query)
};
for(auto it(column.lower_bound(query_key)); bool(it); ++it)
{
const auto key
{
dbs::event_state_key(it->first)
};
const auto &[state_key, type, room_id, depth, event_idx]
{
key
};
const bool tab[]
{
!std::get<0>(query) || std::get<0>(query) == std::get<0>(key),
!std::get<1>(query) || std::get<1>(query) == std::get<1>(key),
!std::get<2>(query) || std::get<2>(query) == std::get<2>(key),
std::get<3>(query) <= 0 || std::get<3>(query) == std::get<3>(key),
std::get<4>(query) == 0 || std::get<4>(query) == std::get<4>(key),
};
if(!std::all_of(begin(tab), end(tab), identity()))
break;
if(!closure(key))
return false;
}
return true;
}
//
// events::type
//
bool
ircd::m::events::type::has(const string_view &type)
{
bool ret{false};
for_each(type, [&ret, &type]
(const string_view &type_)
{
ret = type == type_;
return false; // uncond break out of loop after first result
});
return ret;
}
bool
ircd::m::events::type::has_prefix(const string_view &type)
{
bool ret{false};
for_each(type, [&ret, &type]
(const string_view &type_)
{
ret = startswith(type_, type);
return false; // uncond break out of loop after first result
});
return ret;
}
bool
ircd::m::events::type::for_each_in(const string_view &type,
const closure &closure)
{
auto &column
{
dbs::event_type
};
char buf[dbs::EVENT_TYPE_KEY_MAX_SIZE];
const string_view &key
{
dbs::event_type_key(buf, type)
};
auto it
{
column.begin(key)
};
for(; bool(it); ++it)
{
const auto &keyp
{
dbs::event_type_key(it->first)
};
if(!closure(type, std::get<0>(keyp)))
return false;
}
return true;
}
bool
ircd::m::events::type::for_each(const string_view &prefix,
const closure_name &closure)
{
db::column &column
{
dbs::event_type
};
const auto &prefixer
{
dbs::desc::event_type__pfx
};
string_view last;
char lastbuf[event::TYPE_MAX_SIZE];
for(auto it(column.lower_bound(prefix)); bool(it); ++it)
{
const auto &type
{
prefixer.get(it->first)
};
if(type == last)
continue;
if(prefix && !startswith(type, prefix))
break;
last =
{
lastbuf, copy(lastbuf, type)
};
if(!closure(type))
return false;
}
return true;
}
//
// events::origin
//
bool
ircd::m::events::origin::for_each_in(const string_view &origin,
const sender::closure &closure)
{
auto &column
{
dbs::event_sender
};
char buf[dbs::EVENT_SENDER_KEY_MAX_SIZE];
const string_view &key
{
dbs::event_sender_origin_key(buf, origin)
};
auto it
{
column.begin(key)
};
for(; bool(it); ++it)
{
const auto &keyp
{
dbs::event_sender_origin_key(it->first)
};
const user::id::buf user_id
{
std::get<0>(keyp), origin
};
if(!closure(user_id, std::get<1>(keyp)))
return false;
}
return true;
}
bool
ircd::m::events::origin::for_each(const string_view &prefix,
const closure_name &closure)
{
db::column &column
{
dbs::event_sender
};
const auto &prefixer
{
dbs::desc::event_sender__pfx
};
if(unlikely(startswith(prefix, '@')))
throw panic
{
"Prefix argument should be a hostname. It must not start with '@'"
};
string_view last;
char buf[rfc3986::DOMAIN_BUFSIZE];
for(auto it(column.lower_bound(prefix)); bool(it); ++it)
{
if(!m::dbs::is_event_sender_origin_key(it->first))
break;
const string_view &host
{
prefixer.get(it->first)
};
if(host == last)
continue;
if(!startswith(host, prefix))
break;
last = { buf, copy(buf, host) };
if(!closure(host))
return false;
}
return true;
}
//
// events::sender
//
bool
ircd::m::events::sender::for_each_in(const id::user &user,
const closure &closure)
{
auto &column
{
dbs::event_sender
};
char buf[dbs::EVENT_SENDER_KEY_MAX_SIZE];
const string_view &key
{
dbs::event_sender_key(buf, user)
};
auto it
{
column.begin(key)
};
for(; bool(it); ++it)
{
const auto &keyp
{
dbs::event_sender_key(it->first)
};
if(!closure(user, std::get<event::idx>(keyp)))
return false;
}
return true;
}
bool
ircd::m::events::sender::for_each(const string_view &prefix_,
const closure_name &closure)
{
db::column &column
{
dbs::event_sender
};
const auto &prefixer
{
dbs::desc::event_sender__pfx
};
// We MUST query the column with a key starting with '@' here. For a more
// convenient API, if the user did not supply an '@' we prefix it for them.
char prebuf[id::user::buf::SIZE] {"@"};
const string_view prefix
{
!startswith(prefix_, '@')?
strlcat{prebuf, prefix_}:
prefix_
};
m::user::id::buf last;
for(auto it(column.lower_bound(prefix)); bool(it); ++it)
{
// Check if this is an '@' key; otherwise it's in the origin
// keyspace (sharing this column) which we don't want here.
if(!m::dbs::is_event_sender_key(it->first))
break;
// Apply the domain prefixer, since we're iterating as a db::column
// rather than db::domain.
const m::user::id &user_id
{
prefixer.get(it->first)
};
if(user_id == last)
continue;
if(!startswith(user_id, prefix))
break;
if(!closure(user_id))
return false;
last = user_id;
}
return true;
}