Re: Feature Proposal - async variables (like thread local storage from C)
Yes, it is. Thanks for referring me to the right point. Anything I can do
to help advance that proposal?
On Mar 11, 2018 5:40 PM, "Thomas Grainger" wrote:
I'm pretty sure this is Zones
On 11 Mar 2018 15:35, "Guy Margalit" wrote:
> Hey
>
> It's my first attempt to contribute to ECMAScript so would love your
> feedback if I better off change how I go about it.
>
> The problem I want to highlight is that besides argument passing it is
> impossible to have any information associated with a flow of async code as
> any async code will create its own context and will only get the arguments
> passed to it from its caller.
>
> A simple use case for information that is not easy to pass as arguments is
> "transaction ID" where unique IDs are generated at the beginning of a
> transaction, and propagate that ID to any function that is being called as
> part of that transaction.
>
> However as a javascript codebase grows, it becomes hard to introduce new
> project-wide information such as transaction ID when the only mechanism to
> pass such information is function argument passing, and this becomes
> impossible as many times you are using other libraries of code inside the
> code stack, and one cannot change all the code just to add arguments.
>
> Few examples:
>
>- Identify HTTP requests and print that a request ID in every log
>print that was fan-out from that request.
>- Find how many database calls one async flow created.
>- Keep pools of resources (buffers / connections / tokens) per request.
>
> In C/C++ one can use thread local storage (
> https://en.wikipedia.org/wiki/Thread-local_storage) to keep information
> that automatically switches with the threads context switch.
>
> I imagine a similar mechanism, in the spirit of javascript - define a
> variable as `async`:
>
> ```
> async var name = ;
> ```
>
> Async variables will be saved and restored automatically when the context
> is resumed. It should probably have a more precise definition in ECMAScript
> terms.
>
> ```
> async var reqid = '';
>
> http_server.on('request', handler);
>
> async function handler(req, res) => {
>
> // setting async variable concurrently be different requests
> // will retain the generated value for the other calls made by this
> context (sync or async).
> reqid = `REQ-${uuid()}`;
>
> await load_data_from_db(req);
> await send_analytics_info(req);
> send_reply(res);
> });
>
> async function load_data_from_db(req) {
> console.log(`${reqid} load_data_from_db`);
> try {
> ...
> } catch (err) {
> ...
> if (err.code === 'ACCESS_DENIED') {
> console.log(`${reqid} ACCESS_DENIED. this incident will be reported.
> `);
> }
> }
> }
>
> async function send_analytics_info(req) {
> console.log(`${reqid} send_analytics_info`);
> ...
> analytics.send(reqid, ...);
> ...
> }
>
> function send_reply(res) {
> console.log(`${reqid} send_analytics_info`);
> ...
> res.setHeader('reqid', reqid); // for supportability
> res.send(...);
> ...
> }
> ```
>
> Technically speaking, async variables should be replaced whenever there is
> a context switch - either when a sync stack returns to the top stack
> level, or when an async function returns from await.
>
> However, this means that every registration of a function for later, such
> as setTimeout(func), should also keep the async variables which sounds
> difficult.
>
> Would be great to hear if you think this is valuable and feasible, or
> maybe there's a way to design this with less overhead.
>
> Thanks,
> Guy Margalit
>
>
> ___
> es-discuss mailing list
> es-discuss@mozilla.org
> https://mail.mozilla.org/listinfo/es-discuss
>
>
___
es-discuss mailing list
es-discuss@mozilla.org
https://mail.mozilla.org/listinfo/es-discuss
Re: Feature Proposal - async variables (like thread local storage from C)
I'm pretty sure this is Zones
On 11 Mar 2018 15:35, "Guy Margalit" wrote:
> Hey
>
> It's my first attempt to contribute to ECMAScript so would love your
> feedback if I better off change how I go about it.
>
> The problem I want to highlight is that besides argument passing it is
> impossible to have any information associated with a flow of async code as
> any async code will create its own context and will only get the arguments
> passed to it from its caller.
>
> A simple use case for information that is not easy to pass as arguments is
> "transaction ID" where unique IDs are generated at the beginning of a
> transaction, and propagate that ID to any function that is being called as
> part of that transaction.
>
> However as a javascript codebase grows, it becomes hard to introduce new
> project-wide information such as transaction ID when the only mechanism to
> pass such information is function argument passing, and this becomes
> impossible as many times you are using other libraries of code inside the
> code stack, and one cannot change all the code just to add arguments.
>
> Few examples:
>
>- Identify HTTP requests and print that a request ID in every log
>print that was fan-out from that request.
>- Find how many database calls one async flow created.
>- Keep pools of resources (buffers / connections / tokens) per request.
>
> In C/C++ one can use thread local storage (https://en.wikipedia.org/wiki
> /Thread-local_storage) to keep information that automatically switches
> with the threads context switch.
>
> I imagine a similar mechanism, in the spirit of javascript - define a
> variable as `async`:
>
> ```
> async var name = ;
> ```
>
> Async variables will be saved and restored automatically when the context
> is resumed. It should probably have a more precise definition in ECMAScript
> terms.
>
> ```
> async var reqid = '';
>
> http_server.on('request', handler);
>
> async function handler(req, res) => {
>
> // setting async variable concurrently be different requests
> // will retain the generated value for the other calls made by this
> context (sync or async).
> reqid = `REQ-${uuid()}`;
>
> await load_data_from_db(req);
> await send_analytics_info(req);
> send_reply(res);
> });
>
> async function load_data_from_db(req) {
> console.log(`${reqid} load_data_from_db`);
> try {
> ...
> } catch (err) {
> ...
> if (err.code === 'ACCESS_DENIED') {
> console.log(`${reqid} ACCESS_DENIED. this incident will be reported.
> `);
> }
> }
> }
>
> async function send_analytics_info(req) {
> console.log(`${reqid} send_analytics_info`);
> ...
> analytics.send(reqid, ...);
> ...
> }
>
> function send_reply(res) {
> console.log(`${reqid} send_analytics_info`);
> ...
> res.setHeader('reqid', reqid); // for supportability
> res.send(...);
> ...
> }
> ```
>
> Technically speaking, async variables should be replaced whenever there is
> a context switch - either when a sync stack returns to the top stack
> level, or when an async function returns from await.
>
> However, this means that every registration of a function for later, such
> as setTimeout(func), should also keep the async variables which sounds
> difficult.
>
> Would be great to hear if you think this is valuable and feasible, or
> maybe there's a way to design this with less overhead.
>
> Thanks,
> Guy Margalit
>
>
> ___
> es-discuss mailing list
> es-discuss@mozilla.org
> https://mail.mozilla.org/listinfo/es-discuss
>
>
___
es-discuss mailing list
es-discuss@mozilla.org
https://mail.mozilla.org/listinfo/es-discuss
