neovim/test/client/session.lua

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

199 lines
4.7 KiB
Lua
Raw Normal View History

local uv = require('luv')
local MsgpackRpcStream = require('test.client.msgpack_rpc_stream')
local Session = {}
Session.__index = Session
if package.loaded['jit'] then
-- luajit pcall is already coroutine safe
Session.safe_pcall = pcall
else
Session.safe_pcall = require'coxpcall'.pcall
end
local function resume(co, ...)
local status, result = coroutine.resume(co, ...)
if coroutine.status(co) == 'dead' then
if not status then
error(result)
end
return
end
assert(coroutine.status(co) == 'suspended')
result(co)
end
local function coroutine_exec(func, ...)
local args = {...}
local on_complete
if #args > 0 and type(args[#args]) == 'function' then
-- completion callback
on_complete = table.remove(args)
end
resume(coroutine.create(function()
local status, result, flag = Session.safe_pcall(func, unpack(args))
if on_complete then
coroutine.yield(function()
-- run the completion callback on the main thread
on_complete(status, result, flag)
end)
end
end))
end
function Session.new(stream)
return setmetatable({
_msgpack_rpc_stream = MsgpackRpcStream.new(stream),
_pending_messages = {},
_prepare = uv.new_prepare(),
_timer = uv.new_timer(),
_is_running = false
}, Session)
end
function Session:next_message(timeout)
local function on_request(method, args, response)
table.insert(self._pending_messages, {'request', method, args, response})
uv.stop()
end
local function on_notification(method, args)
table.insert(self._pending_messages, {'notification', method, args})
uv.stop()
end
if self._is_running then
error('Event loop already running')
end
if #self._pending_messages > 0 then
return table.remove(self._pending_messages, 1)
end
refactor(map): enhanced implementation, Clean Code™, etc etc This involves two redesigns of the map.c implementations: 1. Change of macro style and code organization The old khash.h and map.c implementation used huge #define blocks with a lot of backslash line continuations. This instead uses the "implementation file" .c.h pattern. Such a file is meant to be included multiple times, with different macros set prior to inclusion as parameters. we already use this pattern e.g. for eval/typval_encode.c.h to implement different typval encoders reusing a similar structure. We can structure this code into two parts. one that only depends on key type and is enough to implement sets, and one which depends on both key and value to implement maps (as a wrapper around sets, with an added value[] array) 2. Separate the main hash buckets from the key / value arrays Change the hack buckets to only contain an index into separate key / value arrays This is a common pattern in modern, state of the art hashmap implementations. Even though this leads to one more allocated array, it is this often is a net reduction of memory consumption. Consider key+value consuming at least 12 bytes per pair. On average, we will have twice as many buckets per item. Thus old implementation: 2*12 = 24 bytes per item New implementation 1*12 + 2*4 = 20 bytes per item And the difference gets bigger with larger items. One might think we have pulled a fast one here, as wouldn't the average size of the new key/value arrays be 1.5 slots per items due to amortized grows? But remember, these arrays are fully dense, and thus the accessed memory, measured in _cache lines_, the unit which actually matters, will be the fully used memory but just rounded up to the nearest cache line boundary. This has some other interesting properties, such as an insert-only set/map will be fully ordered by insert only. Preserving this ordering in face of deletions is more tricky tho. As we currently don't use ordered maps, the "delete" operation maintains compactness of the item arrays in the simplest way by breaking the ordering. It would be possible to implement an order-preserving delete although at some cost, like allowing the items array to become non-dense until the next rehash. Finally, in face of these two major changes, all code used in khash.h has been integrated into map.c and friends. Given the heavy edits it makes no sense to "layer" the code into a vendored and a wrapper part. Rather, the layered cake follows the specialization depth: code shared for all maps, code specialized to a key type (and its equivalence relation), and finally code specialized to value+key type.
2023-05-17 07:08:06 -07:00
-- if closed, only return pending messages
if self.closed then
return nil
end
self:_run(on_request, on_notification, timeout)
return table.remove(self._pending_messages, 1)
end
function Session:notify(method, ...)
self._msgpack_rpc_stream:write(method, {...})
end
function Session:request(method, ...)
local args = {...}
local err, result
if self._is_running then
err, result = self:_yielding_request(method, args)
else
err, result = self:_blocking_request(method, args)
end
if err then
return false, err
end
return true, result
end
function Session:run(request_cb, notification_cb, setup_cb, timeout)
local function on_request(method, args, response)
coroutine_exec(request_cb, method, args, function(status, result, flag)
if status then
response:send(result, flag)
else
response:send(result, true)
end
end)
end
local function on_notification(method, args)
coroutine_exec(notification_cb, method, args)
end
self._is_running = true
if setup_cb then
coroutine_exec(setup_cb)
end
while #self._pending_messages > 0 do
local msg = table.remove(self._pending_messages, 1)
if msg[1] == 'request' then
on_request(msg[2], msg[3], msg[4])
else
on_notification(msg[2], msg[3])
end
end
self:_run(on_request, on_notification, timeout)
self._is_running = false
end
function Session:stop()
uv.stop()
end
function Session:close(signal)
if not self._timer:is_closing() then self._timer:close() end
if not self._prepare:is_closing() then self._prepare:close() end
self._msgpack_rpc_stream:close(signal)
refactor(map): enhanced implementation, Clean Code™, etc etc This involves two redesigns of the map.c implementations: 1. Change of macro style and code organization The old khash.h and map.c implementation used huge #define blocks with a lot of backslash line continuations. This instead uses the "implementation file" .c.h pattern. Such a file is meant to be included multiple times, with different macros set prior to inclusion as parameters. we already use this pattern e.g. for eval/typval_encode.c.h to implement different typval encoders reusing a similar structure. We can structure this code into two parts. one that only depends on key type and is enough to implement sets, and one which depends on both key and value to implement maps (as a wrapper around sets, with an added value[] array) 2. Separate the main hash buckets from the key / value arrays Change the hack buckets to only contain an index into separate key / value arrays This is a common pattern in modern, state of the art hashmap implementations. Even though this leads to one more allocated array, it is this often is a net reduction of memory consumption. Consider key+value consuming at least 12 bytes per pair. On average, we will have twice as many buckets per item. Thus old implementation: 2*12 = 24 bytes per item New implementation 1*12 + 2*4 = 20 bytes per item And the difference gets bigger with larger items. One might think we have pulled a fast one here, as wouldn't the average size of the new key/value arrays be 1.5 slots per items due to amortized grows? But remember, these arrays are fully dense, and thus the accessed memory, measured in _cache lines_, the unit which actually matters, will be the fully used memory but just rounded up to the nearest cache line boundary. This has some other interesting properties, such as an insert-only set/map will be fully ordered by insert only. Preserving this ordering in face of deletions is more tricky tho. As we currently don't use ordered maps, the "delete" operation maintains compactness of the item arrays in the simplest way by breaking the ordering. It would be possible to implement an order-preserving delete although at some cost, like allowing the items array to become non-dense until the next rehash. Finally, in face of these two major changes, all code used in khash.h has been integrated into map.c and friends. Given the heavy edits it makes no sense to "layer" the code into a vendored and a wrapper part. Rather, the layered cake follows the specialization depth: code shared for all maps, code specialized to a key type (and its equivalence relation), and finally code specialized to value+key type.
2023-05-17 07:08:06 -07:00
self.closed = true
end
function Session:_yielding_request(method, args)
return coroutine.yield(function(co)
self._msgpack_rpc_stream:write(method, args, function(err, result)
resume(co, err, result)
end)
end)
end
function Session:_blocking_request(method, args)
local err, result
local function on_request(method_, args_, response)
table.insert(self._pending_messages, {'request', method_, args_, response})
end
local function on_notification(method_, args_)
table.insert(self._pending_messages, {'notification', method_, args_})
end
self._msgpack_rpc_stream:write(method, args, function(e, r)
err = e
result = r
uv.stop()
end)
self:_run(on_request, on_notification)
return (err or self.eof_err), result
end
function Session:_run(request_cb, notification_cb, timeout)
if type(timeout) == 'number' then
self._prepare:start(function()
self._timer:start(timeout, 0, function()
uv.stop()
end)
self._prepare:stop()
end)
end
self._msgpack_rpc_stream:read_start(request_cb, notification_cb, function()
uv.stop()
self.eof_err = {1, "EOF was received from Nvim. Likely the Nvim process crashed."}
end)
uv.run()
self._prepare:stop()
self._timer:stop()
self._msgpack_rpc_stream:read_stop()
end
return Session