From janos.n.hary at gmail.com Sun May 4 09:59:21 2014 From: janos.n.hary at gmail.com (Janos Hary) Date: Sun, 4 May 2014 09:59:21 +0200 Subject: [99s-extend] Gracefully stop Ranch Message-ID: <001b01cf676e$c2ec8e20$48c5aa60$@gmail.com> All! I implemented a protocol as a ranch_protocol. It handles long running sessions. At some point I'd like to stop my server accepting new connections, but allow all open sessions to finish. Then I need to know when all the open sessions (if any) finished. What would be the correct strategy to implement this? Thanks, Janos From paulo.ferraz.oliveira at gmail.com Tue May 20 18:27:58 2014 From: paulo.ferraz.oliveira at gmail.com (Paulo F. Oliveira) Date: Tue, 20 May 2014 17:27:58 +0100 Subject: [99s-extend] REST responses Message-ID: Hello. First of all, thanks for the great work you've done with cowboy. I've been using it with a fait amount of success and I'm a fairly new Erlang developer. I'm mainly interested in the REST "interface" of the application and its way of doing RESTful things, and I like the way you did it (what with all the content_types_provided, service_available, etc. functions). I've tested the way the system reacted to the different Accept, Content-Type, etc. headers and always got very well-opinionated responses (406, 415, ...). A couple of questions remain though (I'm sorry if they've been asked already but I've searched the web for answers and read the available docs and couldn't find them): 1. is it expected that, if I use cowboy_req:reply/2 in a GET handler (coming from content_types_provided), the onresponse/4 hook be called twice? I guess one is due to the reply and the other one due to the workflow of the request, but is there a way to prevent the second execution? 2. if I want to JSON-parse ALL my requests should I a) use the onrequest/1 hook or b) do this on a per-request basis? Because I'd like to reply with a 400 ASAP but keep going if the JSON validates (I'm going to use JSON-schema for validating input); and, if possible, have the JSON-parsed body stored somewhere for future manipulation. 3. I haven't seen examples that made use of the State (from the function returns). When should I use this instead of the Request metadata? I'd like to be able to set a generic error state for a request (either in meta ou State) and that have a "standard" error response be created at a later time (in a unique function, for example - e.g. onresponse/4). 4. is there anything like a catch-all exception handler? I'd like to catch exceptions that occur anywhere so I could log them and analyze them at a later moment. I'm probably abusing the onresponse/onrequest hooks already, so your answers should help me clarify this. Thanks. - Paulo -------------- next part -------------- An HTML attachment was scrubbed... URL: From paulo.ferraz.oliveira at gmail.com Tue May 20 20:32:10 2014 From: paulo.ferraz.oliveira at gmail.com (Paulo F. Oliveira) Date: Tue, 20 May 2014 19:32:10 +0100 Subject: [99s-extend] 202 for POST or PUT Message-ID: Hi. Do you think it should be possible to generate a 202 for a POST or a PUT? Is it something that will be implemented in a future version? Many thanks. - Paulo -------------- next part -------------- An HTML attachment was scrubbed... URL: From essen at ninenines.eu Tue May 20 20:46:02 2014 From: essen at ninenines.eu (=?UTF-8?B?TG/Dr2MgSG9ndWlu?=) Date: Tue, 20 May 2014 20:46:02 +0200 Subject: [99s-extend] REST responses In-Reply-To: References: Message-ID: <537BA2EA.5080108@ninenines.eu> Hi, On 05/20/2014 06:27 PM, Paulo F. Oliveira wrote: > Hello. > > First of all, thanks for the great work you've done with cowboy. I've > been using it with a fait amount of success and I'm a fairly new Erlang > developer. I'm mainly interested in the REST "interface" of the > application and its way of doing RESTful things, and I like the way you > did it (what with all the content_types_provided, service_available, > etc. functions). I've tested the way the system reacted to the different > Accept, Content-Type, etc. headers and always got very well-opinionated > responses (406, 415, ...). > > A couple of questions remain though (I'm sorry if they've been asked > already but I've searched the web for answers and read the available > docs and couldn't find them): > > 1. is it expected that, if I use cowboy_req:reply/2 in a GET handler > (coming from content_types_provided), the onresponse/4 hook be called > twice? I guess one is due to the reply and the other one due to the > workflow of the request, but is there a way to prevent the second execution? If you reply from a callback you must call {halt, Req, State} to stop processing. > 2. if I want to JSON-parse ALL my requests should I a) use the > onrequest/1 hook or b) do this on a per-request basis? Because I'd like > to reply with a 400 ASAP but keep going if the JSON validates (I'm going > to use JSON-schema for validating input); and, if possible, have the > JSON-parsed body stored somewhere for future manipulation. It makes little sense to do it before the accept callback you define. Not only because you will duplicate content-type checks and whatnot, but also because you don't actually win anything from doing this. If you are using JSON, then JSON processing will take infinitely more resources than the REST code. > 3. I haven't seen examples that made use of the State (from the function > returns). When should I use this instead of the Request metadata? I'd > like to be able to set a generic error state for a request (either in > meta ou State) and that have a "standard" error response be created at a > later time (in a unique function, for example - e.g. onresponse/4). State is for the functions within the current module. Look at cowboy_static for an example. > 4. is there anything like a catch-all exception handler? I'd like to > catch exceptions that occur anywhere so I could log them and analyze > them at a later moment. You can add your own error_logger handler, or use something like lager. All errors end up sending a message to error_logger. > I'm probably abusing the onresponse/onrequest hooks already, so your > answers should help me clarify this. Sounds like it! > Thanks. > > - Paulo > > > _______________________________________________ > Extend mailing list > Extend at lists.ninenines.eu > https://lists.ninenines.eu/listinfo/extend > -- Lo?c Hoguin http://ninenines.eu From essen at ninenines.eu Tue May 20 20:47:16 2014 From: essen at ninenines.eu (=?UTF-8?B?TG/Dr2MgSG9ndWlu?=) Date: Tue, 20 May 2014 20:47:16 +0200 Subject: [99s-extend] 202 for POST or PUT In-Reply-To: References: Message-ID: <537BA334.7080607@ninenines.eu> 202 is only well defined for the DELETE method so there's no plan to add something for other methods at this point. You can of course reply directly if needed. On 05/20/2014 08:32 PM, Paulo F. Oliveira wrote: > Hi. > > Do you think it should be possible to generate a 202 for a POST or a > PUT? Is it something that will be implemented in a future version? > > Many thanks. > > - Paulo > > > _______________________________________________ > Extend mailing list > Extend at lists.ninenines.eu > https://lists.ninenines.eu/listinfo/extend > -- Lo?c Hoguin http://ninenines.eu From paulo.ferraz.oliveira at gmail.com Tue May 20 23:41:15 2014 From: paulo.ferraz.oliveira at gmail.com (Paulo F. Oliveira) Date: Tue, 20 May 2014 22:41:15 +0100 Subject: [99s-extend] REST responses Message-ID: Hi, Lo?c. Thanks for having taken the time to reply. In some of my questions I think I didn't explain myself correctly so I'll give it another go. On 20 May 2014 19:46, Lo?c Hoguin wrote: > Hi, > > > On 05/20/2014 06:27 PM, Paulo F. Oliveira wrote: > >> Hello. >> >> First of all, thanks for the great work you've done with cowboy. I've >> been using it with a fait amount of success and I'm a fairly new Erlang >> developer. I'm mainly interested in the REST "interface" of the >> application and its way of doing RESTful things, and I like the way you >> did it (what with all the content_types_provided, service_available, >> etc. functions). I've tested the way the system reacted to the different >> Accept, Content-Type, etc. headers and always got very well-opinionated >> responses (406, 415, ...). >> >> A couple of questions remain though (I'm sorry if they've been asked >> already but I've searched the web for answers and read the available >> docs and couldn't find them): >> >> 1. is it expected that, if I use cowboy_req:reply/2 in a GET handler >> (coming from content_types_provided), the onresponse/4 hook be called >> twice? I guess one is due to the reply and the other one due to the >> workflow of the request, but is there a way to prevent the second >> execution? >> > > If you reply from a callback you must call {halt, Req, State} to stop > processing. Got it! > 2. if I want to JSON-parse ALL my requests should I a) use the >> onrequest/1 hook or b) do this on a per-request basis? Because I'd like >> to reply with a 400 ASAP but keep going if the JSON validates (I'm going >> to use JSON-schema for validating input); and, if possible, have the >> JSON-parsed body stored somewhere for future manipulation. >> > > It makes little sense to do it before the accept callback you define. Not > only because you will duplicate content-type checks and whatnot, but also > because you don't actually win anything from doing this. If you are using > JSON, then JSON processing will take infinitely more resources than the > REST code. OK, I'll probably stick with a "helper" function that'll do this for me and reply in case there are validation errors. I only found the flow diagrams for the requests today after I had sent this message, and they helped a lot. > 3. I haven't seen examples that made use of the State (from the function >> returns). When should I use this instead of the Request metadata? I'd >> like to be able to set a generic error state for a request (either in >> meta ou State) and that have a "standard" error response be created at a >> later time (in a unique function, for example - e.g. onresponse/4). >> > > State is for the functions within the current module. Look at > cowboy_static for an example. State allows me to, well, keep state, for a request "travelling" through functions, right, and I can change it whenever I want just before returning from a function that is executed prior to another one (the only function for which this doesn't seem to make since is the last one cowboy calls before actually replying to the client)? At the same time, so does the request meta, from what I understood from the manual. So what is the difference between one and the other and when would you recommend one or the other. 4. is there anything like a catch-all exception handler? I'd like to >> catch exceptions that occur anywhere so I could log them and analyze >> them at a later moment. >> > > You can add your own error_logger handler, or use something like lager. > All errors end up sending a message to error_logger. I'll do this, thanks. > I'm probably abusing the onresponse/onrequest hooks already, so your >> answers should help me clarify this. >> > > Sounds like it! > > Thanks. >> >> - Paulo >> >> >> _______________________________________________ >> Extend mailing list >> Extend at lists.ninenines.eu >> https://lists.ninenines.eu/listinfo/extend >> >> > -- > Lo?c Hoguin > http://ninenines.eu > Thanks. -------------- next part -------------- An HTML attachment was scrubbed... URL: From essen at ninenines.eu Tue May 20 23:52:21 2014 From: essen at ninenines.eu (=?UTF-8?B?TG/Dr2MgSG9ndWlu?=) Date: Tue, 20 May 2014 23:52:21 +0200 Subject: [99s-extend] REST responses In-Reply-To: References: Message-ID: <537BCE95.1090100@ninenines.eu> > State allows me to, well, keep state, for a request "travelling" through > functions, right, and I can change it whenever I want just before > returning from a function that is executed prior to another one (the > only function for which this doesn't seem to make since is the last one > cowboy calls before actually replying to the client)? At the same time, > so does the request meta, from what I understood from the manual. So > what is the difference between one and the other and when would you > recommend one or the other. They have different purposes. Meta values are additional information about the request. You're not supposed to set them except in special circumstances, either because you have your own custom protocol like cowboy_rest, or because there's no other way to pass state forward. Don't think about it, always use State. -- Lo?c Hoguin http://ninenines.eu