perf: add lua iteration to decrease loops in lua

Move some of the iteration in to loa and access the values by the index
to reduce the number of loops we need todo to get items into teh results
buffer.

Currently the flow is:
  1) Filter and sort the candidates in rust
  2) Convert to a string and pass to lua
  3) Split the string and add them as lines in a buffer in lua

Now the flow is:
  1) Filter and sort the candidates in rust
  2) Loop over an iterator in lua
  3) Pass each item to lua as a pointer by the index

This removes quite a bit of the work that is needed to get the data into
lua as a table. We are first removing the loop that will join the
results vector into one string. Then we will remove the copy of this
string into lua. We will then finally remove the loop to split the
string and create a table from it in lua. All of this ends up in a 12%
speed up.

Output for `./scripts/bench 0.x`

Benchmark 1: HEAD
  Time (mean ± σ):      2.667 s ±  0.065 s    [User: 8.537 s, System: 1.420 s]
  Range (min … max):    2.588 s …  2.767 s    10 runs

Benchmark 2: 0.x
  Time (mean ± σ):      2.337 s ±  0.150 s    [User: 9.564 s, System: 1.648 s]
  Range (min … max):    2.161 s …  2.529 s    10 runs

Summary
  HEAD ran
    1.14 ± 0.08 times faster than 0.x

-------------------------------------
The percentage difference is -12.00%
-------------------------------------
This commit is contained in:
Ade Attwood 2023-10-22 18:02:50 +01:00
parent d707a6e15b
commit 9f9e4a2023
5 changed files with 173 additions and 24 deletions

View file

@ -20,8 +20,31 @@ ffi.cdef [[
char* ivy_cwd();
int ivy_match(const char*, const char*);
char* ivy_files(const char*, const char*);
int ivy_files_iter(const char*, const char*);
int ivy_files_iter_len(int);
char* ivy_files_iter_at(int, int);
void ivy_files_iter_delete(int);
]]
local iter_mt = {
__len = function(self)
return self.length
end,
__index = function(self, index)
-- Pass in our index -1. This will map lua's one based indexing to zero
-- based indexing that we are using in the rust lib.
local item = ffi.string(ivy_c.ivy_files_iter_at(self.id, index - 1))
return { content = item }
end,
__newindex = function(_, _, _)
error("attempt to update a read-only table", 2)
end,
__gc = function(self)
ivy_c.ivy_files_iter_delete(self.id)
end,
}
local libivy = {}
libivy.ivy_init = function(dir)
@ -37,7 +60,12 @@ libivy.ivy_match = function(pattern, text)
end
libivy.ivy_files = function(pattern, base_dir)
return ffi.string(ivy_c.ivy_files(pattern, base_dir))
local iter_id = ivy_c.ivy_files_iter(pattern, base_dir)
local iter_len = ivy_c.ivy_files_iter_len(iter_id)
local iter = { id = iter_id, length = iter_len }
setmetatable(iter, iter_mt)
return iter
end
return libivy

View file

@ -10,18 +10,37 @@ end)
it("should find a dot file", function(t)
local current_dir = libivy.ivy_cwd()
local matches = libivy.ivy_files(".github/workflows/ci.yml", current_dir)
local results = libivy.ivy_files(".github/workflows/ci.yml", current_dir)
local results = {}
for line in string.gmatch(matches, "[^\r\n]+") do
table.insert(results, line)
if results.length ~= 2 then
t.error("Incorrect number of results found " .. results.length)
end
if #results ~= 2 then
t.error "Incorrect number of results"
end
if results[2] ~= ".github/workflows/ci.yml" then
t.error("Invalid matches: " .. results[2])
if results[2].content ~= ".github/workflows/ci.yml" then
t.error("Invalid matches: " .. results[2].content)
end
end)
it("will allow you to access the length via the metatable", function(t)
local current_dir = libivy.ivy_cwd()
local results = libivy.ivy_files(".github/workflows/ci.yml", current_dir)
local mt = getmetatable(results)
if results.length ~= mt.__len(results) then
t.error "The `length` property does not match the __len metamethod"
end
end)
it("will create an iterator", function(t)
local iter = libivy.ivy_files(".github/workflows/ci.yml", libivy.ivy_cwd())
local mt = getmetatable(iter)
if type(mt["__index"]) ~= "function" then
t.error "The iterator does not have an __index metamethod"
end
if type(mt["__len"]) ~= "function" then
t.error "The iterator does not have an __len metamethod"
end
end)

View file

@ -21,6 +21,22 @@ local function string_to_table(lines)
return matches
end
local function get_items_length(items)
local mt = getmetatable(items)
if mt ~= nil and mt.__len ~= nil then
return mt.__len(items)
end
return #items
end
local function call_gc(items)
local mt = getmetatable(items)
if mt ~= nil and mt.__gc ~= nil then
return mt.__gc(items)
end
end
local window = {}
window.index = 0
@ -106,15 +122,17 @@ window.set_items = function(items)
items = string_to_table(items)
end
local items_length = get_items_length(items)
-- TODO(ade): Validate the items are in the correct format. This also need to
-- come with some descriptive messages and possible help.
-- Display no items text if there are no items to dispaly
if #items == 0 then
if items_length == 0 then
items_length = 1
items = { { content = "-- No Items --" } }
end
local items_length = #items
window.index = items_length - 1
for index = 1, items_length do
@ -130,6 +148,8 @@ window.set_items = function(items)
vim.api.nvim_win_set_height(window.window, line_count)
window.update()
call_gc(items)
end
window.destroy = function()

View file

@ -21,3 +21,13 @@ it("can set items", function(t)
window.set_items { { content = "Line one" } }
t.assert_equal("Line one", window.get_current_selection())
end)
it("will set the items when a string is passed in", function(t)
window.initialize()
local items = table.concat({ "One", "Two", "Three" }, "\n")
window.set_items(items)
local lines = table.concat(vim_mock.get_lines()[window.buffer], "\n")
t.assert_equal(items, lines)
end)

View file

@ -9,21 +9,35 @@ use std::os::raw::{c_char, c_int};
use std::sync::Mutex;
use std::sync::OnceLock;
struct Ivy {
pub file_cache: HashMap<String, Vec<String>>,
}
// A store to the singleton instance of the ivy struct. This must not be accessed directly it must
// be use via the Ivy::global() function. Accessing this directly may cause a panic if its been
// initialized correctly.
static INSTANCE: OnceLock<Mutex<Ivy>> = OnceLock::new();
impl Ivy {
pub fn new() -> Self {
Self {
file_cache: HashMap::new(),
}
}
struct Ivy {
// The file cache so we don't have to keep iterating the filesystem. The map key is the root
// directory that has been search and the value an a vector containing all of the files that as
// in the root. The value will be relative from the root.
pub file_cache: HashMap<String, Vec<String>>,
// The sequence number of the last iterator created. This will use as a pointer value to the
// iterator so we can access it though lua and rust without having to copy strings.
pub iter_sequence: i32,
// A store of all the iterators that have been created. The key is the sequence number and the
// value is the vector of matches that were matched in the search.
pub iter_map: HashMap<i32, Vec<CString>>,
}
impl Ivy {
// Get the global instance of the ivy struct. This will initialize the struct if it has not
// initialized yet.
pub fn global() -> &'static Mutex<Ivy> {
INSTANCE.get_or_init(|| Mutex::new(Ivy::new()))
INSTANCE.get_or_init(|| {
Mutex::new(Ivy {
file_cache: HashMap::new(),
iter_sequence: 0,
iter_map: HashMap::new(),
})
})
}
}
@ -75,6 +89,64 @@ pub fn inner_match(pattern: String, text: String) -> i32 {
m.score(text.as_str()) as i32
}
// Create a new iterator that will iterate over all the files in the given directory that match a
// pattern. It will return the pointer to the iterator so it can be retrieve later. The iterator
// can be deleted with `ivy_files_iter_delete`
#[no_mangle]
pub extern "C" fn ivy_files_iter(c_pattern: *const c_char, c_base_dir: *const c_char) -> i32 {
let directory = to_string(c_base_dir);
let pattern = to_string(c_pattern);
let files = get_files(&directory);
let mut ivy = Ivy::global().lock().unwrap();
// Convert the matches into CStrings so we can pass the pointers out while still maintaining
// ownership. If we didn't do this the CString would be dropped and the pointer would be freed
// while its being used externally.
let sorter_options = sorter::Options::new(pattern);
let matches = sorter::sort_strings(sorter_options, files)
.into_iter()
.map(|m| CString::new(m.content.as_str()).unwrap())
.collect::<Vec<CString>>();
ivy.iter_sequence += 1;
let new_sequence = ivy.iter_sequence;
ivy.iter_map.insert(new_sequence, matches);
new_sequence
}
// Delete the iterator with the given id. This will free the memory used by the iterator that was
// created with `ivy_files_iter`
#[no_mangle]
pub extern "C" fn ivy_files_iter_delete(iter_id: i32) {
let mut ivy = Ivy::global().lock().unwrap();
ivy.iter_map.remove(&iter_id);
}
// Returns the length of a given iterator. This will return the number of items that were matched
// when the iterator was created with `ivy_files_iter`
#[no_mangle]
pub extern "C" fn ivy_files_iter_len(iter_id: i32) -> i32 {
let ivy = Ivy::global().lock().unwrap();
let items = ivy.iter_map.get(&iter_id).unwrap();
items.len() as i32
}
// Returns the item at the given index in the iterator. This will return the full match that was
// given in the iterator. This will return a pointer to the string so it can be used in lua.
#[no_mangle]
pub extern "C" fn ivy_files_iter_at(iter_id: i32, index: i32) -> *const c_char {
let ivy = Ivy::global().lock().unwrap();
let items = ivy.iter_map.get(&iter_id).unwrap();
let item = items.get(index as usize).unwrap();
item.as_ptr()
}
#[no_mangle]
pub extern "C" fn ivy_files(c_pattern: *const c_char, c_base_dir: *const c_char) -> *const c_char {
let pattern = to_string(c_pattern);