about summary refs log tree commit diff
path: root/tvix/eval/src/compiler/bindings.rs
AgeCommit message (Collapse)AuthorFilesLines
2024-02-20 r/7571 refactor(tvix/eval): add SourceCode directly into error typesVincent Ambo1-13/+13
With this change it's no longer necessary to track the SourceCode struct separately from the evaluation for error reporting: It's just stored directly in the errors. This also ends up resolving an issue in compiler::bindings, where we cloned the Arc containing file references way too often. In fact those clones probably compensate for all additional SourceCode clones during error construction now. Change-Id: Ice93bf161e61f8ea3d48103435e20c53e6aa8c3a Reviewed-on: https://cl.tvl.fyi/c/depot/+/10986 Reviewed-by: flokli <flokli@flokli.de> Tested-by: BuildkiteCI
2024-02-02 r/7467 refactor(tvix/eval): Box Value::StringAspen Smith1-5/+5
NixString is *quite* large - like 80 bytes - because of the extra capacity value for BString and because of the context. We want to keep Value small since we're passing it around a lot, so let's box the NixString inside Value::String to save on some memory, and make cloning ostensibly a little cheaper Change-Id: I343c8b4e7f61dc3dcbbaba4382efb3b3e5bbabb2 Reviewed-on: https://cl.tvl.fyi/c/depot/+/10729 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2023-05-25 r/6205 feat(tvix/eval): unthunk empty lists and attribute setsVincent Ambo1-0/+4
Change-Id: Ie66cb1b163a544d45d113fd0f866286f230b0188 Reviewed-on: https://cl.tvl.fyi/c/depot/+/7960 Tested-by: BuildkiteCI Reviewed-by: grfn <grfn@gws.fyi>
2023-03-13 r/5968 chore(tvix/eval): remove `From<SmolStr> for NixString` instanceVincent Ambo1-5/+5
No longer needed, and in some cases caused some extra work. Change-Id: I64e8e7292573bdc92a9c7a8e470e33f8c526f311 Reviewed-on: https://cl.tvl.fyi/c/depot/+/8152 Reviewed-by: raitobezarius <tvl@lahfa.xyz> Tested-by: BuildkiteCI
2023-02-03 r/5831 refactor(tvix/eval): statically resolve select from constant attrsVincent Ambo1-37/+4
When resolving a select expression (`attrs.name` or `attrs.name or default`), if the set compiles to a constant attribute set (as is most notably the case with `builtins`) we can backtrack and replace that attribute set directly with the compiled value. For something like `builtins.length`, this will directly emit an `OpConstant` that leaves the `length` builtin on the stack. Change-Id: I639654e065a06e8cfcbcacb528c6da7ec9e513ee Reviewed-on: https://cl.tvl.fyi/c/depot/+/7957 Tested-by: BuildkiteCI Reviewed-by: flokli <flokli@flokli.de>
2023-01-21 r/5721 refactor(tvix/eval): administer antidote for poisonAdam Joseph1-9/+17
The codebase contains a lot of complexity and odd roundabout handling for shadowing globals. I'm pretty sure none of this is necessary, and all of it disappears if you simply make the globals part of the ordinary identifier resolution chain, with their own scope up above the root scope. Then the ordinary shadowing routines do the right thing, and no special cases or new terminology are required. This commit does that. Note by tazjin: This commit was originally abandoned when Adam decided not to take away reviewer bandwidth for this at the time (eval was still in a much earlier stage). As we've recently done some significant refactoring of globals initialisation this came up again, and it seems we can easily cover the use-cases of the poison tracking in other ways now, so I've rebased, updated and resurrected the CL. Co-Authored-By: Vincent Ambo <tazjin@tvl.su> Signed-off-by: Adam Joseph <adam@westernsemico.com> Change-Id: Ib3309a47a7b31fa5bf10466bade0d876b76ae462 Reviewed-on: https://cl.tvl.fyi/c/depot/+/7089 Reviewed-by: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI Reviewed-by: flokli <flokli@flokli.de>
2023-01-06 r/5606 feat(tvix/eval): warn on empty let-bindingsVincent Ambo1-0/+3
Change-Id: Ib6ef7ce514abbd3e372dfe9df7137aa36dbda9d4 Reviewed-on: https://cl.tvl.fyi/c/depot/+/7770 Reviewed-by: raitobezarius <tvl@lahfa.xyz> Reviewed-by: flokli <flokli@flokli.de> Tested-by: BuildkiteCI
2023-01-06 r/5605 refactor(tvix/eval): short-circuit on empty attrs in compilerVincent Ambo1-0/+9
This is marginally more efficient and has simpler bytecode. Change-Id: Iad37c9aeef24583e8f696911bcd83d43639f2e36 Reviewed-on: https://cl.tvl.fyi/c/depot/+/7769 Reviewed-by: raitobezarius <tvl@lahfa.xyz> Tested-by: BuildkiteCI Reviewed-by: flokli <flokli@flokli.de>
2023-01-06 r/5604 feat(tvix/eval): warn about empty `inherit`sVincent Ambo1-0/+5
Change-Id: I82bec6fe2210bcb88c46fd2fdf3e26bd613d1c1f Reviewed-on: https://cl.tvl.fyi/c/depot/+/7768 Reviewed-by: flokli <flokli@flokli.de> Reviewed-by: raitobezarius <tvl@lahfa.xyz> Tested-by: BuildkiteCI
2023-01-06 r/5601 refactor(tvix/eval): take owned ast::Expr in Compiler::compileVincent Ambo1-3/+3
This adds a very minimal amount of additional Rc-increments (~1 per compilation), but makes it a lot easier to add an AST-optimising compiler pass without incurring a lot of extra cost. Change-Id: I57208bdfc8882e3ae21c5850e14aa380d3ccea36 Reviewed-on: https://cl.tvl.fyi/c/depot/+/7765 Reviewed-by: flokli <flokli@flokli.de> Tested-by: BuildkiteCI
2022-12-25 r/5486 fix(tvix/eval): fix current clippy warningsVincent Ambo1-2/+1
It's been a while since the last time, so quite a lot of stuff has accumulated here. Change-Id: I0762827c197b30a917ff470fd8ae8f220f6ba247 Reviewed-on: https://cl.tvl.fyi/c/depot/+/7597 Reviewed-by: grfn <grfn@gws.fyi> Autosubmit: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI
2022-10-23 r/5183 fix(tvix/eval): fix condition for useless inherit warningVincent Ambo1-2/+2
The warning needs to consider whether it is occuring inside of a thunk, i.e. the dynamic ancestry chain of lambda contexts must be inspected and not just the current scope. Change-Id: I5cf5482d67a8bbb9f03b0ecee7a62f58754f8e59 Reviewed-on: https://cl.tvl.fyi/c/depot/+/7063 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org> Reviewed-by: grfn <grfn@gws.fyi>
2022-10-18 r/5156 fix(tvix/eval): wrap dynamic resolution in an extra thunkVincent Ambo1-2/+9
Without this change it was possible to cause situations (see the new test) in which a `with`-namespace was forced prematurely. Change-Id: I879ea7763b43edc693feace2c73c890d426fafd3 Reviewed-on: https://cl.tvl.fyi/c/depot/+/7031 Autosubmit: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI Reviewed-by: Adam Joseph <adam@westernsemico.com>
2022-10-08 r/5069 fix(tvix/eval): end scope after compiling legacy let bindingsVincent Ambo1-0/+5
When contrasting the compilation of the desugared version to the "sugared" version, this was the noticeable difference. This fixes b/203. Change-Id: Iae02ffc56e06de1de091b84cdc59d8fe83a17d69 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6898 Reviewed-by: grfn <grfn@gws.fyi> Autosubmit: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI
2022-10-04 r/5034 refactor(tvix/eval): remove unnecessary clones in compilerVincent Ambo1-20/+20
There's basically nothing that needs *ownership* of an AST node (which is just a little box full of references to other things anyways), so we can thread this through as references all the way. Change-Id: I35a1348a50c0e8e07d51dfc18847829379166fbf Reviewed-on: https://cl.tvl.fyi/c/depot/+/6853 Tested-by: BuildkiteCI Reviewed-by: grfn <grfn@gws.fyi>
2022-09-30 r/5004 chore(tvix/eval): remove unused field in TrackedBindingsVincent Ambo1-7/+3
Change-Id: I65e31e9173e4f5bba19cc4e3d45eb4f8bf91b424 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6808 Reviewed-by: grfn <grfn@gws.fyi> Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-30 r/5002 feat(tvix/eval): implement nested keysVincent Ambo1-29/+60
This finishes up the implementation of nested keys after the key insight that the nesting level does not need to be tracked, and instead the attribute iterator can simply be retained inside the structures as is (in an advanced state). With this implementation, when encountering a nested key, the Tvix compiler will first analyse whether there is already a matching binding that can be merged (i.e. a binding that is a literal attribute set), and perform the merge, or otherwise create a new recursive set of bindings in which the entry is inserted with the path iterator advanced beyond the first name component. With this, all the logic simply applies recursively until there are no more nested bindings (i.e. until all iterators are "empty"). Note that this has one (potentially insignificant) deviation from Nix currently: If a non-mergable value is supplied (e.g. `a.b = 1; a = 2;`), Tvix will emit a *runtime* error (whereas it is *parse* time in Nix) as the branch which could statically analyse this is currently unreachable. There's a TODO for this, so we can fix it up later. Change-Id: I53df70e09614ff4281a70b80eac7da3beca12da9 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6806 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-30 r/5001 refactor(tvix/eval): split out AttributeSet::from_ast helperVincent Ambo1-26/+31
Change-Id: Id43dbd06aef14cf01b4901d9b3668d790cd2b5ae Reviewed-on: https://cl.tvl.fyi/c/depot/+/6805 Tested-by: BuildkiteCI Reviewed-by: grfn <grfn@gws.fyi>
2022-09-30 r/5000 chore(tvix/eval): remove `nesting_level` trackingVincent Ambo1-25/+2
This is actually quite useless, as we can just pass `AstChildren<ast::Attr>` around after partially consuming it. Change-Id: If0aefa2b53fc801fced1ae0709bff93966bf19f8 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6804 Tested-by: BuildkiteCI Reviewed-by: grfn <grfn@gws.fyi>
2022-09-29 r/4999 refactor(tvix/eval): clean up representation flip in bindingsVincent Ambo1-37/+32
When encountering a nested binding for the first time, cleanly flip the representation to `Binding::Set` in `Binding::merge` before proceeding with the actual merge. This reduces the number of points where we have to deal with the (soon to be slightly more complex) construction of the nested binding representation. Change-Id: Ifd43aac7b59ebd15a72c3ec512386a5bcf26ec13 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6802 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4998 feat(tvix/eval): (partially) track nesting level of attrsetsVincent Ambo1-29/+82
This adds the scaffolding required for tracking the nesting level (and appropriately skipping the correct amount of attrpath entries when inserting nested sets). In order for all of this to work correctly, we can no longer track `AttrpathValue` directly in the entries vector as rnix does not allow us to construct values of that type - so instead we have to track its inner components. Change-Id: Icb18e105586bf6c247c2e66c302cde5609ad9789 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6801 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4997 feat(tvix/eval): merge attribute sets in bindingsVincent Ambo1-13/+110
This is a significant step towards correctly implemented nested bindings. All attribute sets defined within the same binding scope will now be merged as in Nix, if they use the same key. Change-Id: I13e056693d5e73192280043c6dd93b47d1306ed6 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6800 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4996 feat(tvix/eval): add error kind for unmergeable nested attributesVincent Ambo1-1/+4
Change-Id: Ic5e6d1bf2625c33938360affb0d1a7c922af11bf Reviewed-on: https://cl.tvl.fyi/c/depot/+/6799 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4995 feat(tvix/eval): add scaffolding for merging nested attribute setsVincent Ambo1-25/+101
This sets up the required logic for finding and merging attribute sets into nested bindings if they exist. This is absolutely not complete yet and can, at this commit, probably cause undefined runtime behaviour if nested attributes are specified. The basic idea is that a new helper function on the `TrackedBindings` struct is called with each encountered attribute and determines whether the new entry can be merged into an existing attribute or not. Right now the only effect this has in practice is that a new error becomes available if somebody attempts to cause a merge into an inherited key. Change-Id: Id010df3605055eb1ad7fa65241055889dd21bab0 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6798 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4994 refactor(tvix/eval): emit OpAttrs inside of compile_bindingsVincent Ambo1-5/+5
This needs to move here so that we can reuse compile_bindings for the nested attribute sets we're about to start constructing. Change-Id: Ie83f52f7e1d128886e96a1da47792211fa826f21 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6796 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-29 r/4993 refactor(tvix/eval): introduce `TrackedBindings` structVincent Ambo1-11/+34
This struct will be the key to correctly compiling nested bindings, by having insertions flow through some logic that will attempt to bind attribute-set-like things when encountering them. Change-Id: I8b5b20798de60688f3b6dc4526a460ebb2079f6e Reviewed-on: https://cl.tvl.fyi/c/depot/+/6795 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-29 r/4992 refactor(tvix/eval): compile_recursive_scope -> compile_bindingsVincent Ambo1-4/+4
Change-Id: Iff18d0f84ba2b7a4194797e6c52c55b1c37e419c Reviewed-on: https://cl.tvl.fyi/c/depot/+/6794 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-29 r/4989 refactor(tvix/eval): merge all bindings creation logicVincent Ambo1-135/+6
As of this commit, all three types of bindings scopes are compiled the same way (i.e. compilation of non-recursive attribute sets has been switched over to the new code paths). This sets us up for doing the final implementation of nested attribute sets. HOWEVER, this breaks the existing implementation of nested attributes in non-recursive attribute sets. That implementation is flawed and unworkable in practice, so we need to do this dance to be able to implement it correctly. Change-Id: Iba2545c0d1d6b51f5e1a31a5d005b8d01da546d3 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6782 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-29 r/4988 feat(tvix/eval): implement dynamic keys in recursive attrsVincent Ambo1-29/+33
This wires up the new bindings setup logic to be able to thread through & compile dynamic attributes in recursive attrs. It seems like we don't actually need to retain the phasing of Nix exactly, as we can use the phantom mechanism to declare all locals without making the dynamic ones accessible. Change-Id: Ic2d43dd8fd97d7ccd56d8c6adf2ff97274cd837a Reviewed-on: https://cl.tvl.fyi/c/depot/+/6781 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-29 r/4986 feat(tvix/eval): add KeySlot::Dynamic variant for dynamic keysVincent Ambo1-14/+29
Another slice of the salami, but no functionality changes yet (other than opening a code path that can reach a `todo!()`, but this will be removed soon). Change-Id: I56b4ed323f70754ed1ab27964ee3c99cf3bf3292 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6780 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4985 refactor(tvix/eval): clean up unused attrpath normalisation logicVincent Ambo1-52/+36
The previous way of sanitising dynamic keys is going away as we're slowly introducing the new nested key logic. While touching this stuff, I've also changed all the related string types to SmolStr as that is more sensible for identifiers. Change-Id: If30c74151508719d646d0e68e7d6f62c36f4d23f Reviewed-on: https://cl.tvl.fyi/c/depot/+/6779 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4984 chore(tvix/eval): reflow comments in compiler::bindingsVincent Ambo1-101/+88
Change-Id: I6d74f71ecd671feaec96ee4ff39f218907c517fe Reviewed-on: https://cl.tvl.fyi/c/depot/+/6777 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4983 refactor(tvix/eval): merge inherits logic between all binding kindsVincent Ambo1-80/+8
Removes the `compile_inherit_attrs` logic which was only used for BindingsKind::Attrs (i.e. non-recursive attrs). This brings us a step closer to fully merging all the binding logic into one block that can dispatch based on the kind of bindings (and thus giving us a good point to introduce the final logic for nested bindings). Change-Id: If48d7a9497fc084a5cc03a130c2a7da5e2b8ef0c Reviewed-on: https://cl.tvl.fyi/c/depot/+/6776 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-29 r/4982 refactor(tvix/eval): Factor out `declare_bindings` helperVincent Ambo1-41/+63
This helper is responsible for declaring the bindings in the compiler's scope tracking structures. It is almost equivalent to the previous logic, but also accounts for `BindingsKind::Attrs` - though those code paths do not end up here yet. Change-Id: I44f617b99b10f2a7b9675f7b23e2c803a4a93d29 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6775 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-28 r/4979 refactor(tvix/eval): Factor out `bind_values` helperVincent Ambo1-47/+55
This is responsible for actually setting up `TrackedBinding`s on the stack, i.e. in some sense "actually compiling" values in bindings. There is no functionality change to before, i.e. this is a salami slice. Change-Id: Idb0312038e004470a7d130c020ae0fe87c55c218 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6774 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-28 r/4978 refactor(tvix/eval): split `compile_inherits` into twoVincent Ambo1-12/+29
Splits the large `compile_inherits` function which previously *compiled* plain inherits and *declared* namespaced inherits into `compile_plain_inherits` and `declare_namespaced_inherits`. This is supposed to make more sense than before, but is still not consistently used (notably, non-recursive attribute sets still duplicate most of this logic). Another salami slice. Change-Id: Id97fac1cbd5ee97b24d047e7728655e6b7734153 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6773 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-28 r/4977 refactor(tvix/eval): add non-recursive logic to `compile_inherit`Vincent Ambo1-16/+39
... but do not use it yet. This refactoring is pretty complicated, so I'm applying salami-slicing tactics here. Change-Id: I66e04ee10548f68bf67dc842f3f14cc279426c22 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6772 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-28 r/4976 refactor(tvix/eval): introduce type to track kind of bindingsVincent Ambo1-11/+30
As part of the unification of binding logic between different carriers of bindings, we need to track which kind of bindings we are dealing with (attribute set? recursive scope? ...) to correctly emit keys and declare identifiers in the locals stack. Right now this changes no functionality as `BindingsKind::Attrs` is not yet used (only RecAttrs and LetIn, which was previously represented by the `rec_attrs` boolean). Change-Id: Id2ac27894079ab584521cb568d75c124f7bf2403 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6771 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-28 r/4975 refactor(tvix/eval): move recursive inherit logic into helperVincent Ambo1-98/+108
This helper will gain the ability to compile both kinds of inherits, but it is kind of tricky to get right so I am doing it in smaller steps. Right now there is no change in functionality. Change-Id: Ie990b88dd90a5e0f9fd79961ee09a6c83f2c872d Reviewed-on: https://cl.tvl.fyi/c/depot/+/6770 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-28 r/4974 refactor(tvix/eval): rename BindingKind -> BindingVincent Ambo1-7/+7
This just describes a binding, and we do need a good name for the kind of binding*s*, which is going to be introduced soon. Change-Id: I53900ee52da8a07dae8b918fa6a4cb308e627efb Reviewed-on: https://cl.tvl.fyi/c/depot/+/6768 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-28 r/4973 refactor(tvix/eval): generalise error variant for dynamic keysVincent Ambo1-26/+12
Change-Id: I08f40b4b53652a519e76d6e8344c7c3fe10a0689 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6767 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-28 r/4972 refactor(tvix/eval): bye compiler::attrs, hello compiler::bindingsVincent Ambo1-0/+689
Changes the module structure of the compiler to have a module dedicated to the logic of setting up bindings. This logic is in the process of being merged between attribute sets and `let`-expressions, and the structure of the modules makes more sense when ecapsulating that specifically. (Other bits of code related to e.g. attribute sets are pretty straightforward and can just live in the main compiler module). Change-Id: I9469b73a7034e5b5f3bb211694d97260c4c9ef54 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6766 Autosubmit: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
generated by cgit-pink 1.4.1 (git 2.44.2) at 2024-11-02T01·41+0000