[0:02] when chatting with someone on Jupiter's
[0:03] thirdd largest moon you will notice a
[0:06] distinct lag in Communications that is
[0:08] due to async
[0:14] IO when you try to perform a task that
[0:16] requires a lot of waiting you may notice
[0:19] an annoying lag in execution you can
[0:21] solve this with async IO this module
[0:25] brings the await and async syntax to
[0:27] python this pair of felet words empowers
[0:30] you with the tools to easily manage a
[0:33] collection of laggy
[0:34] tasks sometimes impatience is a
[0:40] virtue code can be categorized as
[0:43] synchronous or
[0:44] asynchronous in synchronous programming
[0:47] tasks are performed one at a time in the
[0:49] order they are called each task must
[0:52] finish before the next one can
[0:54] begin with asynchronous programming
[0:56] however different tasks can start
[0:59] process and finish in overlapping
[1:01] periods of time for example imagine a
[1:04] task that submits many HTTP requests in
[1:08] the synchronous World these HTTP
[1:11] requests must be completed one at a time
[1:14] in the asynchronous world each request
[1:17] can start and be set aside while waiting
[1:19] for a response during this waiting
[1:21] period different tasks can step forward
[1:24] and use the Computing resources this
[1:27] nonlinear approach saves one of the most
[1:29] Val resources around time to better
[1:34] understand the mechanics of synchronous
[1:35] and asynchronous execution you need to
[1:37] learn about sub routines and co-routines
[1:40] a sub routine is a block of code that
[1:42] can be called as needed when a sub
[1:44] routine is called control of the program
[1:47] is transferred to that sub routine when
[1:49] its work is done control is returned
[1:52] back to the main program and execution
[1:54] continues from where it left off sub
[1:57] routines cannot be paused and resumed
[2:00] they run until
[2:02] done a co- routine is a special kind of
[2:05] function that can have its executions
[2:07] paused and resumed this is possible
[2:10] because they maintain their state
[2:12] between pauses co- routines are perfect
[2:15] for tasks that need to wait for
[2:17] something like IO operations database
[2:20] calls and HTTP requests the terms sub
[2:24] routine and co-routine are derived from
[2:27] the roles they play and how they
[2:28] function sub routine is a delightful
[2:31] blend of sub and routine as the name
[2:35] suggests it is a subset of a larger
[2:37] program co- routine is a ju toos of Co
[2:41] and routine Co for together and routine
[2:46] for
[2:47] routine this name was chosen by someone
[2:49] to describe routines that can run
[2:52] cooperatively co-routines are the key
[2:54] ingredient to make asynchronous
[2:56] programming or concurrency possible in
[2:59] in a typical single-threaded app all of
[3:02] the code and sub routines run
[3:04] sequentially this is simple but can be
[3:07] wasteful there are two different
[3:08] approaches that make better use of
[3:10] available Computing resources
[3:12] concurrency and
[3:14] parallelism using concurrency the start
[3:17] and stop times of multiple co- routines
[3:20] can overlap with parallelism different
[3:23] threads can execute at the same time is
[3:27] there going to be any python in this
[3:28] python lesson
[3:30] yes yes there is but before we can enter
[3:34] python mode I must import an important
[3:37] message socratica is now in the course
[3:39] business visit socratic.com to see the
[3:42] excitement for yourself and for an
[3:45] asynchronous update about future courses
[3:47] feverishly type your email address in a
[3:50] text box somewhere on the site now back
[3:54] to python we are now going to show you
[3:56] how to use concurrency in Python with
[3:58] the async and await keywords and the
[4:01] async io
[4:03] module let us create two functions brew
[4:08] coffee and toast
[4:12] Bagel The Brew Coffee function will use
[4:14] the sleep function to simulate a command
[4:17] that takes 3 seconds to complete the
[4:19] toast Bagel function will pretend it
[4:21] takes 2 seconds to complete with this
[4:24] 5-second toasty breakfast the future is
[4:26] fast and
[4:28] carbolicious we will record a start time
[4:31] call both
[4:33] functions record a stop
[4:35] time compute the elaps time and print
[4:38] the
[4:44] results
[4:50] run this code took 5 seconds and a bit
[4:53] more to execute but there is no reason
[4:56] to wait for the coffee to finish before
[4:58] you start toasting the B Bagel such
[5:00] inefficiency gives the secretly embedded
[5:03] AI systems far too much time for
[5:05] Mischief let us modify this code so it
[5:08] runs
[5:09] concurrently with
[5:11] co-routines to turn each function into a
[5:14] co- routine add the async keyword this
[5:18] is the first step but not the last as is
[5:21] evident if you run using the async
[5:24] keyword to make a co- routine is
[5:26] necessary but not sufficient you also
[5:29] need to to specify where in the co-
[5:30] routine it is safe to pause and yield
[5:33] control to other co- routines you do
[5:35] this with the await
[5:38] keyword this brings us to another
[5:40] essential concept you can only put await
[5:43] in front of commands that are awaitable
[5:46] the sleep function is not
[5:48] awaitable luckily the async io module
[5:51] gives you an awaitable version of the
[5:53] sleep
[5:54] function you can call the co- routines
[5:57] individually or in a batch we will first
[6:00] demonstrate the batch approach the async
[6:03] io module has a gather function to group
[6:05] co- routines for concurrent execution
[6:08] the arguments determine which co-
[6:10] routines will run
[6:11] concurrently the arguments may appear to
[6:13] be regular function calls but they are
[6:15] not calling a co-routine returns a
[6:18] co-routine object not a typical return
[6:21] value these co-routine objects give
[6:24] async IO the ability to start and stop
[6:27] their
[6:28] execution to get the return values from
[6:30] the co- routines you need to await them
[6:34] and note that you await the return
[6:35] values in the same order you pass the
[6:37] co- routines to The Gather function
[6:40] order matters and the use of a weight
[6:43] here is not optional after all we have
[6:47] to wait for both co- routines to finish
[6:50] so it only makes sense that we have to
[6:51] wait on the batch too we are so close to
[6:54] the finish line but there is one final
[6:57] step see how we await for the batch
[7:00] results any function that has an await
[7:02] keyword must be declared async this
[7:06] means we must put the async keyword
[7:08] before the main
[7:09] function because main has now become a
[7:12] co-routine we have to call it in a
[7:14] slightly different way the async io
[7:17] module has you covered just call the Run
[7:20] function and pass in the main co-
[7:22] routine now inhale widen your eyes and
[7:28] run
[7:32] Joy of Joys our breakfast was completed
[7:35] in 3 seconds instead of five time saved
[7:39] me
[7:41] happy there is an alternative to using
[7:43] the gather function you can use the
[7:46] create task function in the async io
[7:48] module to create a task out of each co-
[7:52] routine then you can await each co-
[7:54] routine
[7:57] individually run
[8:03] just as fast just as happy so which is
[8:08] better working in batches or with
[8:10] individual tasks it is your choice so
[8:14] Choose Wisely computer chips want to
[8:17] compute that is their singular purpose
[8:19] in life the async io module was built to
[8:23] help chips realize their full potential
[8:26] by adding a pause button to objects you
[8:28] enable them to take turns sharing is
[8:31] caring after all and speaking of sharing
[8:34] many human people ask you to share their
[8:36] videos with as many human people as
[8:38] possible this is not an efficient use of
[8:41] resources instead await the opportunity
[8:44] to yield this video to someone who has
[8:47] finished a task and is thinking about
[8:50] the
[8:54] [Music]
[8:58] future
[9:02] [Music]