golang / go

The Go programming language
https://go.dev
BSD 3-Clause "New" or "Revised" License
123.8k stars 17.64k forks source link

net: clarify safe way to do URL deep-copying #38351

Open rasky opened 4 years ago

rasky commented 4 years ago

It sometimes happens to have a need to duplicate (aka "deep copy") a net.URL object.

My understanding is that net.URL can be safely duplicated by simply copying the structure:

   newUrl := *url

Since net.URL is usually handled by pointer, it's normal to check the documentation to see whether the object contains internal pointers, in which case duplication might get hairy. net.URL has the following contents:

type URL struct {
    Scheme     string
    Opaque     string    // encoded opaque data
    User       *Userinfo // username and password information
    Host       string    // host or host:port
    Path       string    // path (relative paths may omit leading slash)
    RawPath    string    // encoded path hint (see EscapedPath method); added in Go 1.5
    ForceQuery bool      // append a query ('?') even if RawQuery is empty; added in Go 1.7
    RawQuery   string    // encoded query values, without '?'
    Fragment   string    // fragment for references, without '#'
}

So there is indeed a pointer in there. This brought people to come up with alternative ways of duplicating a URL such as a String/Parse roundtrip: https://medium.com/@nrxr/quick-dirty-way-to-deep-copy-a-url-url-on-go-464312241de4

The trick here is that User *Userinfo is actually "immutable". This is documented in Userinfo: The Userinfo type is an immutable encapsulation of username and password [...]. So it looks like that, given this guarantee of immutability, it is indeed safe to simply copy a URL instance to duplicate it. Copying a URL instance is also performed by the standard library, for instance:

https://github.com/golang/go/blob/245409ea86f20fd9f4167223c2339fb238f9e4b6/src/net/http/request.go#L361

Hence the subject of this issue:

abemotion commented 4 years ago
User       *Userinfo // username and password immutable information

This documentation is better?

nrxr commented 4 years ago

Hi @rasky,

Indeed it's safe and actually I, the author of that Medium post, used the newU := *u method in a PR that got merged some weeks ago: https://github.com/golang/go/pull/35578/files#diff-6c2d018290e298803c0c9419d8739885R824-R837

icholy commented 4 years ago

Relevant comment https://github.com/golang/go/issues/41733#issuecomment-718048068 by @rsc

I still don't understand what exactly is worth calling out in URL's documentation. It is a general property of data that if you want to make a copy before mutating you can do u2 := *u; mutate u2.

And I still don't understand how often this operation is needed on URLs. @ainar-g has seen it, but that establishes existence not frequency.

realitycheck commented 2 years ago

There is something confusing about #41733 due to there actually is cloneURL function in net/http internals not just a u2 := *u thing, however it's not a part of url.URL public protocol itself for some reason.

paralin commented 7 months ago

Ran into this recently where there is no good way to deep clone a URL.