Hello, I’m writing an auth module according to this: https://github.com/vladfaust/crystalworld/blob/master/src/endpoints/auth.cr Instead of checking the header, it needs to check the request body. However, in the endpoint that includes this module a
json macro is used, and it seems the context.request.body is already consumed by it before the before callback is called. Any ideas?
Hey, @qszhu, thanks for the question! You could try rewinding the request body (
context.request.body.rewind), but I'm not sure if it would work 🤔. For some reason in current release there is no way to preserve the body... Another workaround is to copy the body in a handler (i.e. custom middleware)
rewind is unimplemented since the body I’m getting is a FixedLengthContent < IO::Sized: https://github.com/crystal-lang/crystal/blob/b42631e9b3c62621f42832755427d530c6e452d2/src/http/common.cr#L37 I tried to patch HTTP::Request like this: qszhu/http@d15c225 It works, but it feels awkward (and possibly with additional memory impact?). I’m still new to crystal, so any suggestion is welcome.
@qszhu yes, body is always of fixed length. Copying the body on every request (i.e. monkey-patching the
HTTP::Request class) would indeed affect the performance. That's why I've opened onyxframework/http#77. Note that the body is read on json and form params parsing only, and it should stay unread in an endpoint if no such a parsing happened
Well, if I read it correctly, the body is only parsed if the content length is not zero. Thus, it’s only copied in those POST/PUT requests that you usually need to do json/form params parsing. The only difference is when there is a request body and you chose not to parse params from it. Am I right? Still, copying is not ideal. I used to implement this kind of stuff in a custom middleware plugged into specific routes (in Express.js/Koa.js). It seems there’s no easier way than mangling with the fixed middleware pipeline of onyx right now.
@aemadrid try this: https://gist.github.com/qszhu/b31fa84f47a4c17581ff1dc21daa7049 .
It’s undocumented so I think it’s better not to rely on it.
It’s undocumented so I think it’s better not to rely on it.
but it is still a nice way to document/show people what your app can do
figured out my problem but it would have been nice to know
Hey @vladfaust quick question, i was just testing EDA with a redis url that contains a password. it resulted in this error
DEBUG [19:56:36.550 #1] [redis] CLIENT ID
Unhandled exception: NOAUTH Authentication required. (MiniRedis::Error)
from ???
from /usr/share/crystal/src/pointer.cr:434:13 in 'receive'
from lib/mini_redis/src/mini_redis.cr:176:15 in 'send_impl'
from lib/mini_redis/src/mini_redis.cr:89:5 in '???'
from lib/onyx-http/src/onyx-http/ext/http/request/path_params.cr:1:1 in '__crystal_main'
from /usr/share/crystal/src/crystal/main.cr:47:14 in 'main'
from __libc_start_main
from _start
from ???
Well, there is no specification in Redis for URI authentication, and mini_redis does not implement any custom wrapping as well
As a workaround, you could pass custom MiniRedis instance to Onyx::EDA:
Onyx::EDA::Channel::Redis.new
Just call
redis.send("AUTH", password) before, as seen at redis.io
You could create an issue @ mini_redis if you think this should be automated
As most of query builder methods support raw SQL
I never intended to replace SQL. You should know raw SQL to make it work better, anyway
I’m trying to return all the fields of a reference with join according to the doc: https://docs.onyxframework.org/sql/query.html#join Say I want to get all the fields of author. Do I have to write raw sql like
.join(author: true) do |x|
x.select(“author.*”)
endor is there any other way?
Hey @qszhu, sorry for the late response. You can do this:
.join(author: true, &.select(User))This would effectively select all the fields of the author with proper table naming author. instead of users.
I am not really getting how to make this work.
(first time I am trying to use a database instead of flat files)
what I would like to do is:
p headings[0].book.titleThe code...
(sqlite)
CREATE TABLE book (
book INTEGER PRIMARY KEY,
title TEXT NOT NULL
);
CREATE TABLE heading (
heading INTEGER PRIMARY KEY,
title TEXT NOT NULL,
page INTEGER NOT NULL,
book INTEGER NOT NULL,
FOREIGN KEY (book) REFERENCES book(book)
);class Models::Book
include Onyx::SQL::Model
schema book do
pkey book : Int64
type title : String, not_null: true
end
end
class Models::Heading
include Onyx::SQL::Model
schema heading do
pkey heading : Int64
type title : String, not_null: true
type page : Int32, not_null: true
type book : Book, not_null: true, key: "book"
end
end
headings = Onyx::SQL.query(Models::Heading.join(book: true, &.select(Models::Book)))
pp headingsThe error I get when trying the above
Error in src/mini_index.cr:73: no overload matches 'Models::Heading.join', book: Bool
Overloads are:
- Onyx::SQL::Model::ClassQueryShortcuts(T)#join(table : String, on : String, as _as : String | ::Nil = nil, type : Onyx::SQL::Query::JoinType = :inner)
- Onyx::SQL::Model::ClassQueryShortcuts(T)#join(reference : T::Reference, on : String | ::Nil = nil, as _as : String = reference.to_s.underscore, type : Onyx::SQL::Query::JoinType = :inner)
- Onyx::SQL::Model::ClassQueryShortcuts(T)#join(reference : T::Reference, klass, *, on : String | ::Nil = nil, as _as : String = reference.to_s.underscore, type : JoinType = :inner, &block)
headings = Onyx::SQL.query(Models::Heading.join(book: true, &.select(Models::Book)))I can work around it like this
headings = Onyx::SQL.query(Models::Heading.all)
books = Onyx::SQL.query(Models::Book.all)
pp books.select {|b| b.book == headings[0].book.try(&.book)} .first.titleDoes anyone have any idea of what I am doing wrong?
Can I use EDA https://github.com/onyxframework/eda to emit a generic Event so I can handle events from "all" events?
Then when I have
Foo(Specific) i can emit it, then make a new Foo(All) from that and emit it too. Not ideal but it's okay
anyone has a workaround ?