Hi, in OSX I cannot get the CodeGenerator to run from Unity - I get the ".Net Core SDK not found" error even though I did install it. I set a break point when it is testing for it by executing "dotnet --version" and the error that is returned is "Cannot find the specified file". I noticed the PATH environment variable did not include the location of the dotnet executable so I manually edited to add /usr/local/share/dotnet, but I still get the same error.
this is a necessary step unless I'm using the Typeless formatter right?
I am running into the same problem with Dictionary<string,object> dict; and other types, e.g. Decimal. I would expect that the ContractlessStandardResolver.Instance does pack the type information and I receive a decimal if i deserialize dict["key"] = 14m; but I do receive a string.
resolver = CompositeResolver.Create(
NativeGuidResolver.Instance,
NativeDecimalResolver.Instance,
DynamicObjectResolver.Instance,
ContractlessStandardResolver.Instance,
StandardResolver.Instance
);
... ah this is still won't fix neuecc/MessagePack-CSharp#651 but - sorry I thought it was a "won't" fix in v1
do you also happen to know whether the new docs has the latest benchmark against version 2 on it or not?
this issue was happenning before if I was using the mpc file without the .dll extension, the latest mpc doesn’t come with dll, so I have no choice to use it with mono, or let the rider detect that it’s to be run/built against dotnet and not mono (as it auto detects based on the file type)
Another option might be to prefix with
cmd or bash or whatever so you can control the whole thing, as these are not .NET executables either.
@AArnott I can’t prefix it with anything as the config asks for certain fields only to construct the command. Rider uses
mono by default as the main cmd tool if the file is either .exe or doesn’t have any extension, unless it comes with .dll which then it uses dotnet instead
why can’t we just have the mpc.dll any more?
You can file an issue on our site to track shipping a .dll
But surely you can pass parameters into the tool from a Rider config, can't you?
If so, can't you point Rider at a native .exe with args?
I have spoken to Rider guys, so far they suggested to use the
Native Executable template than .Net Executable, though it doesn’t justfiy the fact that Rider defaults to mono for any .net executable instead letting the user to choose between dotnet and mono. But there’s caveat of how to use it which gives the same result using a normal console as well, tha path arguments for input project file (.csproj) and the output file (generated file) need to be absolute path and not relative even if running under the working directory, otherwise it internally fails with System.ArgumentException: The path is empty. (Parameter ‘path’)! This issue could also be related to neuecc/MessagePack-CSharp#355 as well
Hi all. I have been using MP in it's native for for many years and love it! First time I needed to do something special would like your advise. We have our own set of attributes that we use on objects that we mark for various purposes. One of them is whether member needs to be serialize. Without adding extra MP attributes such as [Key] or [IgnoreMember] how would I approach this problem? I also do not want to create a proxy class to add attributes on the fly dynamically.
The messagepack resolver that looks for attributes directly turns them to a dynamic formatter. There is no mechanism in place to get it to read other attributes or allow you to direct it yourself after reading your own attributes. I'd love to see this enhanced in the future, but that's where things are now.
@neuecc what do you think of me adding a GitHub Action (bot) that tags stale issues and PRs and then closes them if no activity happens? There would be an 'exempt' label you could apply on long-term issues, and we could negotiate the 'stale' timeframe. I'm thinking 90 days is stale, +5 days to close.
First run is complete. It only tagged 15 issues since it has a 30 operation limit by default. So each day it will get some of the job done. We could lift the limit, but since this is brand new, maybe doing a small batch each day is a good idea so we can keep tabs on how it goes.
@provanguard If NSwag can produce
[DataContract] conforming classes to represent the data to be serialized, then yes it should work fine.
@AArnott Andrew. I wave looked for docs on how to write custom Formatter resolver and seemed to be missing something. Can you please point me in the right direction as to how to write a custom forrmatter resolver? While I have used MP a lot, I am not familiar with its internals. Just need an example please.
@AArnott I have*
@AArnott After initial investigation, it seem that NSwag / NSwagStudio focuses (solely?) on JSON serialization. It is a mystery to me why binary serialization is not a option. I am investigating this to find a fast and reliable inter-service communication. The goal is to have fast response from chained services and binary serialization looks very promising. In that context, generation of client code would be very handy.