flatbuffers/tests/luatest.lua

429 lines
12 KiB
Lua
Raw Normal View History

Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
package.path = string.format("../lua/?.lua;./?.lua;%s",package.path)
local compat = require("flatbuffers.compat")
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local performBenchmarkTests = false
if #arg > 1 then
print("usage: lua luatests [benchmark]");
return
elseif #arg > 0 then
if(arg[1] == "benchmark") then
performBenchmarkTests = true
end
end
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local function checkReadBuffer(buf, offset, sizePrefix)
offset = offset or 0
if type(buf) == "string" then
buf = flatbuffers.binaryArray.New(buf)
end
if sizePrefix then
local size = flatbuffers.N.Int32:Unpack(buf, offset)
assert(size == buf.size - offset - 4)
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
offset = offset + flatbuffers.N.Int32.bytewidth
end
local mon = monster.GetRootAsMonster(buf, offset)
assert(mon:Hp() == 80, "Monster Hp is not 80")
assert(mon:Mana() == 150, "Monster Mana is not 150")
assert(mon:Name() == "MyMonster", "Monster Name is not MyMonster")
assert(mon:Testbool() == true)
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local vec = assert(mon:Pos(), "Monster Position is nil")
assert(vec:X() == 1.0)
assert(vec:Y() == 2.0)
assert(vec:Z() == 3.0)
assert(vec:Test1() == 3.0)
assert(vec:Test2() == 2)
local t = require("MyGame.Example.Test").New()
t = assert(vec:Test3(t))
assert(t:A() == 5)
assert(t:B() == 6)
local ut = require("MyGame.Example.Any")
assert(mon:TestType() == ut.Monster)
local table2 = mon:Test()
assert(getmetatable(table2) == "flatbuffers.view.mt")
local mon2 = monster.New()
mon2:Init(table2.bytes, table2.pos)
assert(mon2:Name() == "Fred")
assert(mon:InventoryLength() == 5)
local invsum = 0
for i=1,mon:InventoryLength() do
local v = mon:Inventory(i)
invsum = invsum + v
end
assert(invsum == 10)
for i=1,5 do
assert(mon:VectorOfLongs(i) == 10^((i-1)*2))
end
local dbls = { -1.7976931348623157e+308, 0, 1.7976931348623157e+308}
for i=1,mon:VectorOfDoublesLength() do
assert(mon:VectorOfDoubles(i) == dbls[i])
end
assert(mon:Test4Length() == 2)
local test0 = mon:Test4(1)
local test1 = mon:Test4(2)
local v0 = test0:A()
local v1 = test0:B()
local v2 = test1:A()
local v3 = test1:B()
local sumtest12 = v0 + v1 + v2 + v3
assert(sumtest12 == 100)
assert(mon:TestarrayofstringLength() == 2)
assert(mon:Testarrayofstring(1) == "test1")
assert(mon:Testarrayofstring(2) == "test2")
assert(mon:TestarrayoftablesLength() == 0)
assert(mon:TestnestedflatbufferLength() == 0)
assert(mon:Testempty() == nil)
end
local function generateMonster(sizePrefix, b)
2019-11-14 23:46:47 +00:00
if b then b:Clear() end
b = b or flatbuffers.Builder(0)
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local str = b:CreateString("MyMonster")
local test1 = b:CreateString("test1")
local test2 = b:CreateString("test2")
local fred = b:CreateString("Fred")
monster.StartInventoryVector(b, 5)
b:PrependByte(4)
b:PrependByte(3)
b:PrependByte(2)
b:PrependByte(1)
b:PrependByte(0)
local inv = b:EndVector(5)
monster.Start(b)
monster.AddName(b, fred)
local mon2 = monster.End(b)
monster.StartTest4Vector(b, 2)
test.CreateTest(b, 10, 20)
test.CreateTest(b, 30, 40)
local test4 = b:EndVector(2)
monster.StartTestarrayofstringVector(b, 2)
b:PrependUOffsetTRelative(test2)
b:PrependUOffsetTRelative(test1)
local testArrayOfString = b:EndVector(2)
monster.StartVectorOfLongsVector(b, 5)
b:PrependInt64(100000000)
b:PrependInt64(1000000)
b:PrependInt64(10000)
b:PrependInt64(100)
b:PrependInt64(1)
local vectorOfLongs = b:EndVector(5)
monster.StartVectorOfDoublesVector(b, 3)
b:PrependFloat64(1.7976931348623157e+308)
b:PrependFloat64(0)
b:PrependFloat64(-1.7976931348623157e+308)
local vectorOfDoubles = b:EndVector(3)
monster.Start(b)
local pos = vec3.CreateVec3(b, 1.0, 2.0, 3.0, 3.0, 2, 5, 6)
monster.AddPos(b, pos)
monster.AddHp(b, 80)
monster.AddName(b, str)
monster.AddInventory(b, inv)
monster.AddTestType(b, 1)
monster.AddTest(b, mon2)
monster.AddTest4(b, test4)
monster.AddTestbool(b, true)
monster.AddTestbool(b, false)
monster.AddTestbool(b, null)
monster.AddTestbool(b,"true")
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
monster.AddTestarrayofstring(b, testArrayOfString)
monster.AddVectorOfLongs(b, vectorOfLongs)
monster.AddVectorOfDoubles(b, vectorOfDoubles)
local mon = monster.End(b)
if sizePrefix then
b:FinishSizePrefixed(mon)
else
b:Finish(mon)
end
return b:Output(true), b:Head()
end
local function sizePrefix(sizePrefix)
local buf,offset = generateMonster(sizePrefix)
checkReadBuffer(buf, offset, sizePrefix)
end
local function fbbClear()
-- Generate a builder that will be 'cleared' and reused to create two different objects.
local fbb = flatbuffers.Builder(0)
-- First use the builder to read the normal monster data and verify it works
local buf, offset = generateMonster(false, fbb)
checkReadBuffer(buf, offset, false)
-- Then clear the builder to be used again
fbb:Clear()
-- Storage for the built monsters
local monsters = {}
local lastBuf
-- Make another builder that will be use identically to the 'cleared' one so outputs can be compared. Build both the
-- Cleared builder and new builder in the exact same way, so we can compare their results
for i, builder in ipairs({fbb, flatbuffers.Builder(0)}) do
local strOffset = builder:CreateString("Hi there")
monster.Start(builder)
monster.AddPos(builder, vec3.CreateVec3(builder, 3.0, 2.0, 1.0, 17.0, 3, 100, 123))
monster.AddName(builder, strOffset)
monster.AddMana(builder, 123)
builder:Finish(monster.End(builder))
local buf = builder:Output(false)
if not lastBuf then
lastBuf = buf
else
-- the output, sized-buffer should be identical
assert(lastBuf == buf, "Monster output buffers are not identical")
end
monsters[i] = monster.GetRootAsMonster(flatbuffers.binaryArray.New(buf), 0)
end
-- Check that all the fields for the generated monsters are as we expect
for i, monster in ipairs(monsters) do
assert(monster:Name() == "Hi there", "Monster Name is not 'Hi There' for monster "..i)
-- HP is default to 100 in the schema, but we change it in generateMonster to 80, so this is a good test to
-- see if the cleared builder really clears the data.
assert(monster:Hp() == 100, "HP doesn't equal the default value for monster "..i)
assert(monster:Mana() == 123, "Monster Mana is not '123' for monster "..i)
assert(monster:Pos():X() == 3.0, "Monster vec3.X is not '3' for monster "..i)
end
end
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local function testCanonicalData()
local f = assert(io.open('monsterdata_test.mon', 'rb'))
local wireData = f:read("*a")
f:close()
checkReadBuffer(wireData)
end
local function testCreateEmptyString()
local b = flatbuffers.Builder(0)
local str = b:CreateString("")
monster.Start(b)
monster.AddName(b, str)
b:Finish(monster.End(b))
local s = b:Output()
local data = flatbuffers.binaryArray.New(s)
local mon = monster.GetRootAsMonster(data, 0)
assert(mon:Name() == "")
end
2019-11-14 23:46:47 +00:00
local function benchmarkMakeMonster(count, reuseBuilder)
local fbb = reuseBuilder and flatbuffers.Builder(0)
local length = #(generateMonster(false, fbb))
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local s = os.clock()
for i=1,count do
2019-11-14 23:46:47 +00:00
generateMonster(false, fbb)
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
end
local e = os.clock()
local dur = (e - s)
local rate = count / (dur * 1000)
local data = (length * count) / (1024 * 1024)
local dataRate = data / dur
print(string.format('built %d %d-byte flatbuffers in %.2fsec: %.2f/msec, %.2fMB/sec',
count, length, dur, rate, dataRate))
end
local function benchmarkReadBuffer(count)
local f = assert(io.open('monsterdata_test.mon', 'rb'))
local buf = f:read("*a")
f:close()
local s = os.clock()
for i=1,count do
checkReadBuffer(buf)
end
local e = os.clock()
local dur = (e - s)
local rate = count / (dur * 1000)
local data = (#buf * count) / (1024 * 1024)
local dataRate = data / dur
print(string.format('traversed %d %d-byte flatbuffers in %.2fsec: %.2f/msec, %.2fMB/sec',
count, #buf, dur, rate, dataRate))
end
local function getRootAs_canAcceptString()
local f = assert(io.open('monsterdata_test.mon', 'rb'))
local wireData = f:read("*a")
f:close()
assert(type(wireData) == "string", "Data is not a string");
local mon = monster.GetRootAsMonster(wireData, 0)
assert(mon:Hp() == 80, "Monster Hp is not 80")
end
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local function testAccessByteVectorAsString()
local f = assert(io.open('monsterdata_test.mon', 'rb'))
local wireData = f:read("*a")
f:close()
local mon = monster.GetRootAsMonster(wireData, 0)
-- the data of byte array Inventory is [0, 1, 2, 3, 4]
local s = mon:InventoryAsString(1, 3)
assert(#s == 3)
for i = 1, #s do
assert(string.byte(s, i) == i - 1)
end
local s = mon:InventoryAsString(2, 5)
assert(#s == 4)
for i = 1, #s do
assert(string.byte(s, i) == i)
end
local s = mon:InventoryAsString(5, 5)
assert(#s == 1)
assert(string.byte(s, 1) == 4)
local s = mon:InventoryAsString(2)
assert(#s == 4)
for i = 1, #s do
assert(string.byte(s, i) == i)
end
local s = mon:InventoryAsString()
assert(#s == 5)
for i = 1, #s do
assert(string.byte(s, i) == i - 1)
end
end
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local tests =
{
{
f = sizePrefix,
d = "Test size prefix",
args = {{true}, {false}}
},
{
f = fbbClear,
d = "FlatBufferBuilder Clear",
},
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
{
f = testCanonicalData,
d = "Tests Canonical flatbuffer file included in repo"
},
{
f = testCreateEmptyString,
d = "Avoid infinite loop when creating empty string"
},
{
f = getRootAs_canAcceptString,
d = "Tests that GetRootAs<type>() generated methods accept strings"
},
{
f = testAccessByteVectorAsString,
d = "Access byte vector as string"
},
}
local benchmarks =
{
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
{
f = benchmarkMakeMonster,
d = "Benchmark making monsters",
args = {
{100},
{1000},
{10000},
2019-11-14 23:46:47 +00:00
{10000, true}
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
}
},
{
f = benchmarkReadBuffer,
d = "Benchmark reading monsters",
args = {
{100},
{1000},
{10000},
-- uncomment following to run 1 million to compare.
-- Took ~141 seconds on my machine
2019-11-14 23:46:47 +00:00
--{1000000},
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
}
},
}
local result, err = xpcall(function()
flatbuffers = assert(require("flatbuffers"))
monster = assert(require("MyGame.Example.Monster"))
test = assert(require("MyGame.Example.Test"))
vec3 = assert(require("MyGame.Example.Vec3"))
local function buildArgList(tbl)
local s = ""
for _,item in ipairs(tbl) do
s = s .. tostring(item) .. ","
end
return s:sub(1,-2)
end
if performBenchmarkTests then
for _,benchmark in ipairs(benchmarks) do
table.insert(tests, benchmark)
end
end
Lua (5.3) Language addition (#4804) * starting Lua port of python implmention. Syncing commit * Bulk of Lua module port from Python done. Not tested, only static analysis. Need to work on binary strings. Started work on flatc lua code generation * Fixed all the basic errors to produced a binary output from the builder, don't know if it is generated correctly, but it contains data, so that must be good * fixed binary set command that was extending the array improperly * continued improvement * Moved lua submodules down a directory so their names don't clash with potential other modules. Added compat module to provide Lua versioning logic * Successful sample port from Python * working on testing Lua code with formal tests * continued to work on tests and fixes to code to make tests pass * Added reading buffer test * Changed binaryarray implmentation to use a temporary table for storing data, and then serialize it to a string when requested. This double the rate of building flatbuffers compared to the string approach. * Didn't need encode module as it just added another layer of indirection that isn't need * profiled reading buffers, optimizations to increase read performance of monster data to ~7 monster / millisecond * Writing profiler improvments. Get about ~2 monsters/millisecond building rate * removed Numpy generation from Lua (came from the Python port) * math.pow is deprecated in Lua 5.3, so changed to ^ notation. Also added .bat script for starting Lua tests * adding results of generate_code.bat * simple edits for code review in PR. * There was a buffer overflow in inserting the keywords into the unorder set for both the Lua and Python code gens. Changed insertion to use iterators. * fixed spacing issue * basic documenation/tutorial updates. Updated sample_binary.lua to reflect the tutorial better * removed windows-specific build step in Lua tests
2018-07-05 22:55:57 +00:00
local testsPassed, testsFailed = 0,0
for _,test in ipairs(tests) do
local allargs = test.args or {{}}
for _,args in ipairs(allargs) do
local results, err = xpcall(test.f,debug.traceback, table.unpack(args))
if results then
testsPassed = testsPassed + 1
else
testsFailed = testsFailed + 1
print(string.format(" Test [%s](%s) failed: \n\t%s",
test.d or "",
buildArgList(args),
err))
end
end
end
local totalTests = testsPassed + testsFailed
print(string.format("# of test passed: %d / %d (%.2f%%)",
testsPassed,
totalTests,
totalTests ~= 0
and 100 * (testsPassed / totalTests)
or 0)
)
return 0
end, debug.traceback)
if not result then
print("Unable to run tests due to test framework error: ",err)
end
os.exit(result and 0 or -1)