[RFC 0045] Deprecating unquoted URL syntax #45
Conversation
rfcs/0045-deprecate-url-syntax.md
Outdated
| strings or uniformly unquoted. | ||
|
|
||
| Decide to use unquoted URLs for all URLs without special characters or variable | ||
| expansion. |
There was a problem hiding this comment.
Mind making these two items a list?
There was a problem hiding this comment.
Definitely done in the PR branch, I have no idea why GitHub doesn't show this yet…
rfcs/0045-deprecate-url-syntax.md
Outdated
| convert URLs to quoted strings when changing them. | ||
|
|
||
| Accept PRs that convert unquoted URLs to quoted strings if such PRs are | ||
| submitted. |
There was a problem hiding this comment.
I wonder if there is a way ofborg could detect newly introduced URL values?
There was a problem hiding this comment.
Maybe your profiling patch to Nix is in the right direction for that…
|
Strongly in favour. It's never been clear to me why this was part of the syntax, and more than once I've been left with analysis paralysis because I wasn't sure whether there was a reason to use the special syntax or not, and there appeared to be no documentation on that. And from a tooling implementor POV: removing this in the long run would reduce the complexity of a Nix (parser) implementation. |
|
Quick thoughts about deprecating language/interpreter features: As far as detection goes, deprecating features in the interpreter could be made strict with a list of "new" features, e.g. Though, this is all at the cost of adding more conditionals, thus more possibilities for weird behaviours. Gating it in the main interpreter with an option would also allow end-users to set it globally, allowing new deprecation-free evals to be forced on their systems, hopefully helping weed out the leftover features deprecations. |
globin
changed the title
globin
changed the title
|
Thanks for the suggestion of one more reason, let's have motivation section be the longest one! |
rfcs/0045-deprecate-url-syntax.md
Outdated
|
|
||
| Add a note in the Nixpkgs manual that the unquoted URL syntax is deprecated, | ||
| changes to Nixpkgs should not increase its use, and it is recommended to | ||
| convert URLs to quoted strings when changing them. |
There was a problem hiding this comment.
Is it possible to be more specific? How would everybody be notified about this new rule?
There was a problem hiding this comment.
- I think the mere fact of an RFC passing is a noticeable effect event for a large fraction of active contributors. (It is visible in this repository, announced via Discourse, gets into the «weekly», is likely to be discussed a bit on IRC etc.)
For this specific RFC just the announcement «RFC on Deprecation of URL syntax passed» conveys enough information.
-
It will be noted in two out of three manuals.
-
Once we have an RFC that something «should not» happen, a request for cosmetic cleanup in a PR (and such requests happen) is more likely to contain a mention of quoting URLs (if relevant for the package in question). This also spreads the knowledge.
I do have an impression that quite a few people mentioned that they are disappointed they cannot refer to a policy that quoted URLs are better, so I expect the review channel of information dissemination to perform well.
Appendix A: future work — maybe the tool gets implemented, then we can open a countdown issue and maybe make ofborg check that PRs do not make things worse).
My plan is indeed just these points. I think that these things do not really need any additions to the text of RFC, and I think these mechanisms will be enough to distribute the information.
If you think that any of these four channels benefit from an addition to RFC text, or that there are other information distribution channels that should be used (and mentioned in the RFC text) please give some details.
There was a problem hiding this comment.
It doesn't have to be formalized in the RFC but it's good to have an accompanying discussion.
For example you mentioned "they are disappointed they cannot refer to a policy". Would the RFC act as such policy? If not, maybe we should have a list of official policies in place.
There was a problem hiding this comment.
I think having a deprecation notice in the manual (added in accordance with an explicit bullet point in an approved/accepted RFC) is close enough to official policy for the purpose of asking people not to do the deprecated things.
I didn't start the discussion myself, because I thought that (1) is assumed, (2) and (A) are explicitly mentioned, and (3) is a manual variant of (A) in a sense.
| start-date: 2019-04-28 | ||
| author: Michael Raskin | ||
| co-authors: | ||
| related-issues: |
There was a problem hiding this comment.
There was a problem hiding this comment.
I am not sure: the RFC template seems to imply this is for PRs implementing the RFC, not for the changes the RFC intends to undo.
Co-Authored-By: 7c6f434c <7c6f434c@mail.ru>
Co-Authored-By: 7c6f434c <7c6f434c@mail.ru>
|
@FRidh thanks for the proofreading. |
|
We should not break backwards compatibility in such a massive way. I mean, deprecation is only useful if you eventually want to remove it, at which point it becomes impossible for Nix to build old Nix expressions. In principle, the flakes "epoch" feature would make it possible to make such language changes in a backward compatible way (e.g. unquoted URLs could be deprecated in epoch 2020). I'm not convinced it's actually useful, though. The case of |
|
There are some features that we have chosen not to use in Nixpkgs anymore and we do not recommend to use these features if they can be avoided at all. Additional tooling intended for aiding Nixpkgs development can be useful without supporting such features, and easier creation of tooling is good. The RFC definitely does not argue for actually dropping the support unless some other breaking changes happen. It will be a pity if an epoch-guarded syntax change for something major like a change in overrides happens without a cleanup of minor annnoyances. |
|
In 2022 we can parse Nix identifier without first parsing a URI with a regex, which will speed up the parser. |
|
@domenkozar I think URIs, identifiers etc. are all recognized using a single finite state machine, so it shouldn't matter much for performance. |
Mic92
added
the
status: open for nominations
unquoted urls are deprecated: NixOS/rfcs#45
url literals were deprecated by rfc 45 NixOS/rfcs#45
url literals were deprecated by rfc 45 NixOS/rfcs#45
|
This pull request has been mentioned on NixOS Discourse. There might be relevant details there: |
Basically an attempt to resume fixing NixOS#5543 for a breakage introduced earlier[1]. Basically, when evaluating an older `nixpkgs` with `nix-shell` the following error occurs: λ ma27 [~] → nix-shell -I nixpkgs=channel:nixos-18.03 -p nix error: anonymous function at /nix/store/zakqwc529rb6xcj8pwixjsxscvlx9fbi-source/pkgs/top-level/default.nix:20:1 called with unexpected argument 'inNixShell' at /nix/store/zakqwc529rb6xcj8pwixjsxscvlx9fbi-source/pkgs/top-level/impure.nix:82:1: 81| 82| import ./. (builtins.removeAttrs args [ "system" "platform" ] // { | ^ 83| inherit config overlays crossSystem; This is a problem because one of the main selling points of Nix is that you can evaluate any old Nix expression and still get the same result (which also means that it *still evaluates*). In fact we're deprecating, but not removing a lot of stuff for that reason such as unquoted URLs[2] or `builtins.toPath`. However this property was essentially thrown away here. The change is rather simple: check if `inNixShell` is specified in the formals of an auto-called function. This means that { inNixShell ? false }: builtins.trace inNixShell (with import <nixpkgs> { }; makeShell { name = "foo"; }) will show `trace: true` while args@{ ... }: builtins.trace args.inNixShell (with import <nixpkgs> { }; makeShell { name = "foo"; }) will throw the following error: error: attribute 'inNixShell' missing This is explicitly needed because the function in `pkgs/top-level/impure.nix` of e.g. NixOS 18.03 has an ellipsis[3], but passes the attribute-set on to another lambda with formals that doesn't have an ellipsis anymore (hence the error from above). This was perhaps a mistake, but we can't fix it anymore. This also means that there's AFAICS no proper way to check if the attr-set that's passed to the Nix code via `EvalState::autoCallFunction` is eventually passed to a lambda with formals where `inNixShell` is missing. However, this fix comes with a certain price. Essentially every `shell.nix` that assumes `inNixShell` to be passed to the formals even without explicitly specifying it would break with this[4]. However I think that this is ugly, but preferable: * Nix 2.3 was declared stable by NixOS up until recently (well, it still is as long as 21.11 is alive), so most people might not have even noticed that feature. * We're talking about a way shorter time-span with this change being in the wild, so the fallout should be smaller IMHO. [1] NixOS@9d612c3 [2] NixOS/rfcs#45 (comment) [3] https://github.com/NixOS/nixpkgs/blob/release-18.03/pkgs/top-level/impure.nix#L75 [4] See e.g. the second expression in this commit-message or the changes for `tests/ca/nix-shell.sh`.
Basically an attempt to resume fixing NixOS#5543 for a breakage introduced earlier[1]. Basically, when evaluating an older `nixpkgs` with `nix-shell` the following error occurs: λ ma27 [~] → nix-shell -I nixpkgs=channel:nixos-18.03 -p nix error: anonymous function at /nix/store/zakqwc529rb6xcj8pwixjsxscvlx9fbi-source/pkgs/top-level/default.nix:20:1 called with unexpected argument 'inNixShell' at /nix/store/zakqwc529rb6xcj8pwixjsxscvlx9fbi-source/pkgs/top-level/impure.nix:82:1: 81| 82| import ./. (builtins.removeAttrs args [ "system" "platform" ] // { | ^ 83| inherit config overlays crossSystem; This is a problem because one of the main selling points of Nix is that you can evaluate any old Nix expression and still get the same result (which also means that it *still evaluates*). In fact we're deprecating, but not removing a lot of stuff for that reason such as unquoted URLs[2] or `builtins.toPath`. However this property was essentially thrown away here. The change is rather simple: check if `inNixShell` is specified in the formals of an auto-called function. This means that { inNixShell ? false }: builtins.trace inNixShell (with import <nixpkgs> { }; makeShell { name = "foo"; }) will show `trace: true` while args@{ ... }: builtins.trace args.inNixShell (with import <nixpkgs> { }; makeShell { name = "foo"; }) will throw the following error: error: attribute 'inNixShell' missing This is explicitly needed because the function in `pkgs/top-level/impure.nix` of e.g. NixOS 18.03 has an ellipsis[3], but passes the attribute-set on to another lambda with formals that doesn't have an ellipsis anymore (hence the error from above). This was perhaps a mistake, but we can't fix it anymore. This also means that there's AFAICS no proper way to check if the attr-set that's passed to the Nix code via `EvalState::autoCallFunction` is eventually passed to a lambda with formals where `inNixShell` is missing. However, this fix comes with a certain price. Essentially every `shell.nix` that assumes `inNixShell` to be passed to the formals even without explicitly specifying it would break with this[4]. However I think that this is ugly, but preferable: * Nix 2.3 was declared stable by NixOS up until recently (well, it still is as long as 21.11 is alive), so most people might not have even noticed that feature. * We're talking about a way shorter time-span with this change being in the wild, so the fallout should be smaller IMHO. [1] NixOS@9d612c3 [2] NixOS/rfcs#45 (comment) [3] https://github.com/NixOS/nixpkgs/blob/release-18.03/pkgs/top-level/impure.nix#L75 [4] See e.g. the second expression in this commit-message or the changes for `tests/ca/nix-shell.sh`.
Unquoted URLs are deprecated NixOS/rfcs#45
Unquoted URLs are deprecated NixOS/rfcs#45
otherwise, eval fails when the experimental no-url-literals feature is activated unquoted urls are discouraged after NixOS/rfcs#45
No description provided.