after #6218 `Symbol` no longer confers a uniqueness invariant on the
string it wraps, it is now possible to create multiple symbols that
compare equal but whose string contents have different addresses. this
guarantee is now only provided by `SymbolIdx`, leaving `Symbol` only as
a string wrapper that knows about the intricacies of how symbols need to
be formatted for output.
this change renames `SymbolIdx` to `Symbol` to restore the previous
semantics of `Symbol` to that name. we also keep the wrapper type and
rename it to `SymbolStr` instead of returning plain strings from lookups
into the symbol table because symbols are formatted for output in many
places. theoretically we do not need `SymbolStr`, only a function that
formats a string for output as a symbol, but having to wrap every symbol
that appears in a message into eg `formatSymbol()` is error-prone and
inconvient.
this slightly increases the amount of memory used for any given symbol, but this
increase is more than made up for if the symbol is referenced more than once in
the EvalState that holds it. on average every symbol should be referenced at
least twice (once to introduce a binding, once to use it), so we expect no
increase in memory on average.
symbol tables are limited to 2³² entries like position tables, and similar
arguments apply to why overflow is not likely: 2³² symbols would require as many
string instances (at 24 bytes each) and map entries (at 24 bytes or more each,
assuming that the map holds on average at most one item per bucket as the docs
say). a full symbol table would require at least 192GB of memory just for
symbols, which is well out of reach. (an ofborg eval of nixpks today creates
less than a million symbols!)
PosTable deduplicates origin information, so using symbols for paths is no
longer necessary. moving away from path Symbols also reduces the usage of
symbols for things that are not keys in attribute sets, which will become
important in the future when we turn symbols into indices as well.
Pos objects are somewhat wasteful as they duplicate the origin file name and
input type for each object. on files that produce more than one Pos when parsed
this a sizeable waste of memory (one pointer per Pos). the same goes for
ptr<Pos> on 64 bit machines: parsing enough source to require 8 bytes to locate
a position would need at least 8GB of input and 64GB of expression memory. it's
not likely that we'll hit that any time soon, so we can use a uint32_t index to
locate positions instead.
the only use of this function is to determine whether a lambda has a non-set
formal, but this use is arguably better served by Symbol::set and using a
non-Symbol instead of an empty symbol in the parser when no such formal is present.
if we defer the duplicate argument check for lambda formals we can use more
efficient data structures for the formals set, and we can get rid of the
duplication of formals names to boot. instead of a list of formals we've seen
and a set of names we'll keep a vector instead and run a sort+dupcheck step
before moving the parsed formals into a newly created lambda. this improves
performance on search and rebuild by ~1%, pure parsing gains more (about 4%).
this does reorder lambda arguments in the xml output, but the output is still
stable. this shouldn't be a problem since argument order is not semantically
important anyway.
before
nix search --no-eval-cache --offline ../nixpkgs hello
Time (mean ± σ): 8.550 s ± 0.060 s [User: 6.470 s, System: 1.664 s]
Range (min … max): 8.435 s … 8.666 s 20 runs
nix eval -f ../nixpkgs/pkgs/development/haskell-modules/hackage-packages.nix
Time (mean ± σ): 346.7 ms ± 2.1 ms [User: 312.4 ms, System: 34.2 ms]
Range (min … max): 343.8 ms … 353.4 ms 20 runs
nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.720 s ± 0.031 s [User: 2.415 s, System: 0.231 s]
Range (min … max): 2.662 s … 2.780 s 20 runs
after
nix search --no-eval-cache --offline ../nixpkgs hello
Time (mean ± σ): 8.462 s ± 0.063 s [User: 6.398 s, System: 1.661 s]
Range (min … max): 8.339 s … 8.542 s 20 runs
nix eval -f ../nixpkgs/pkgs/development/haskell-modules/hackage-packages.nix
Time (mean ± σ): 329.1 ms ± 1.4 ms [User: 296.8 ms, System: 32.3 ms]
Range (min … max): 326.1 ms … 330.8 ms 20 runs
nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.687 s ± 0.035 s [User: 2.392 s, System: 0.228 s]
Range (min … max): 2.626 s … 2.754 s 20 runs
string expressions by and large do not need the benefits a Symbol gives us,
instead they pollute the symbol table and cause unnecessary overhead for almost
all strings. the one place we can think of that benefits from them (attrpaths
with expressions) extracts the benefit in the parser, which we'll have to touch
anyway when changing ExprString to hold strings.
this gives a sizeable improvement on of 3-5% on all benchmarks we've run.
before
nix search --no-eval-cache --offline ../nixpkgs hello
Time (mean ± σ): 8.844 s ± 0.045 s [User: 6.750 s, System: 1.663 s]
Range (min … max): 8.758 s … 8.922 s 20 runs
nix eval -f ../nixpkgs/pkgs/development/haskell-modules/hackage-packages.nix
Time (mean ± σ): 367.4 ms ± 3.3 ms [User: 332.3 ms, System: 35.2 ms]
Range (min … max): 364.0 ms … 375.2 ms 20 runs
nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.810 s ± 0.030 s [User: 2.517 s, System: 0.225 s]
Range (min … max): 2.742 s … 2.854 s 20 runs
after
nix search --no-eval-cache --offline ../nixpkgs hello
Time (mean ± σ): 8.533 s ± 0.068 s [User: 6.485 s, System: 1.642 s]
Range (min … max): 8.404 s … 8.657 s 20 runs
nix eval -f ../nixpkgs/pkgs/development/haskell-modules/hackage-packages.nix
Time (mean ± σ): 347.6 ms ± 3.1 ms [User: 313.1 ms, System: 34.5 ms]
Range (min … max): 343.3 ms … 354.6 ms 20 runs
nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.709 s ± 0.032 s [User: 2.414 s, System: 0.232 s]
Range (min … max): 2.655 s … 2.788 s 20 runs
it can be replaced with StringToken if we add another bit if information to
StringToken, namely whether this string should take part in indentation scanning
or not. since all escaping terminates indentation scanning we need to set this
bit only for the non-escaped IND_STRING rule.
this improves performance by about 1%.
before
nix search --no-eval-cache --offline ../nixpkgs hello
Time (mean ± σ): 8.880 s ± 0.048 s [User: 6.809 s, System: 1.643 s]
Range (min … max): 8.781 s … 8.993 s 20 runs
nix eval -f ../nixpkgs/pkgs/development/haskell-modules/hackage-packages.nix
Time (mean ± σ): 375.0 ms ± 2.2 ms [User: 339.8 ms, System: 35.2 ms]
Range (min … max): 371.5 ms … 379.3 ms 20 runs
nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.831 s ± 0.040 s [User: 2.536 s, System: 0.225 s]
Range (min … max): 2.769 s … 2.912 s 20 runs
after
nix search --no-eval-cache --offline ../nixpkgs hello
Time (mean ± σ): 8.832 s ± 0.048 s [User: 6.757 s, System: 1.657 s]
Range (min … max): 8.743 s … 8.921 s 20 runs
nix eval -f ../nixpkgs/pkgs/development/haskell-modules/hackage-packages.nix
Time (mean ± σ): 367.4 ms ± 3.2 ms [User: 332.7 ms, System: 34.7 ms]
Range (min … max): 364.6 ms … 374.6 ms 20 runs
nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.810 s ± 0.030 s [User: 2.517 s, System: 0.225 s]
Range (min … max): 2.742 s … 2.854 s 20 runs
when given a string yacc will copy the entire input to a newly allocated
location so that it can add a second terminating NUL byte. since the
parser is a very internal thing to EvalState we can ensure that having
two terminating NUL bytes is always possible without copying, and have
the parser itself merely check that the expected NULs are present.
# before
Benchmark 1: nix search --offline nixpkgs hello
Time (mean ± σ): 572.4 ms ± 2.3 ms [User: 563.4 ms, System: 8.6 ms]
Range (min … max): 566.9 ms … 579.1 ms 50 runs
Benchmark 2: nix eval -f ../nixpkgs/pkgs/development/haskell-modules/hackage-packages.nix
Time (mean ± σ): 381.7 ms ± 1.0 ms [User: 348.3 ms, System: 33.1 ms]
Range (min … max): 380.2 ms … 387.7 ms 50 runs
Benchmark 3: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.936 s ± 0.005 s [User: 2.715 s, System: 0.221 s]
Range (min … max): 2.923 s … 2.946 s 50 runs
# after
Benchmark 1: nix search --offline nixpkgs hello
Time (mean ± σ): 571.7 ms ± 2.4 ms [User: 563.3 ms, System: 8.0 ms]
Range (min … max): 566.7 ms … 579.7 ms 50 runs
Benchmark 2: nix eval -f ../nixpkgs/pkgs/development/haskell-modules/hackage-packages.nix
Time (mean ± σ): 376.6 ms ± 1.0 ms [User: 345.8 ms, System: 30.5 ms]
Range (min … max): 374.5 ms … 379.1 ms 50 runs
Benchmark 3: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.922 s ± 0.006 s [User: 2.707 s, System: 0.215 s]
Range (min … max): 2.906 s … 2.934 s 50 runs
every stringy token the lexer returns is turned into a Symbol and not
used further, so we don't have to strdup. using a string_view is
sufficient, but due to limitations of the current parser we have to use
a POD type that holds the same information.
gives ~2% on system build, 6% on search, 8% on parsing alone
# before
Benchmark 1: nix search --offline nixpkgs hello
Time (mean ± σ): 610.6 ms ± 2.4 ms [User: 602.5 ms, System: 7.8 ms]
Range (min … max): 606.6 ms … 617.3 ms 50 runs
Benchmark 2: nix eval -f hackage-packages.nix
Time (mean ± σ): 430.1 ms ± 1.4 ms [User: 393.1 ms, System: 36.7 ms]
Range (min … max): 428.2 ms … 434.2 ms 50 runs
Benchmark 3: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 3.032 s ± 0.005 s [User: 2.808 s, System: 0.223 s]
Range (min … max): 3.023 s … 3.041 s 50 runs
# after
Benchmark 1: nix search --offline nixpkgs hello
Time (mean ± σ): 574.7 ms ± 2.8 ms [User: 566.3 ms, System: 8.0 ms]
Range (min … max): 569.2 ms … 580.7 ms 50 runs
Benchmark 2: nix eval -f hackage-packages.nix
Time (mean ± σ): 394.4 ms ± 0.8 ms [User: 361.8 ms, System: 32.3 ms]
Range (min … max): 392.7 ms … 395.7 ms 50 runs
Benchmark 3: nix eval --raw --impure --expr 'with import <nixpkgs/nixos> {}; system'
Time (mean ± σ): 2.976 s ± 0.005 s [User: 2.757 s, System: 0.218 s]
Range (min … max): 2.966 s … 2.990 s 50 runs
We now parse function applications as a vector of arguments rather
than as a chain of binary applications, e.g. 'substring 1 2 "foo"' is
parsed as
ExprCall { .fun = <substring>, .args = [ <1>, <2>, <"foo"> ] }
rather than
ExprApp (ExprApp (ExprApp <substring> <1>) <2>) <"foo">
This allows primops to be called immediately (if enough arguments are
supplied) without having to allocate intermediate tPrimOpApp values.
On
$ nix-instantiate --dry-run '<nixpkgs/nixos/release-combined.nix>' -A nixos.tests.simple.x86_64-linux
this gives a substantial performance improvement:
user CPU time: median = 0.9209 mean = 0.9218 stddev = 0.0073 min = 0.9086 max = 0.9340 [rejected, p=0.00000, Δ=-0.21433±0.00677]
elapsed time: median = 1.0585 mean = 1.0584 stddev = 0.0024 min = 1.0523 max = 1.0623 [rejected, p=0.00000, Δ=-0.20594±0.00236]
because it reduces the number of tPrimOpApp allocations from 551990 to
42534 (i.e. only small minority of primop calls are partially
applied) which in turn reduces time spent in the garbage collector.
Rather than having them plain strings scattered through the whole
codebase, create an enum containing all the known experimental features.
This means that
- Nix can now `warn` when an unkwown experimental feature is passed
(making it much nicer to spot typos and spot deprecated features)
- It’s now easy to remove a feature altogether (once the feature isn’t
experimental anymore or is dropped) by just removing the field for the
enum and letting the compiler point us to all the now invalid usages
of it.
The boolean is only used to determine if the formals are set to a
non-null pointer in all our cases. We can get rid of that allocation and
instead just compare the pointer value with NULL. Saving up to
sizeof(bool) + platform specific alignment per ExprLambda instace.
Probably not a lot of memory but perhaps a few kilobyte with nixpkgs?
This also gets rid of a potential issue with dereferencing formals based on
the value of the boolean that didn't have to be aligned with the formals
pointer but was in all our cases.
Previously, type or coercion errors for string interpolation, path
interpolation, and plus expressions were always reported at the
beginning of the outer expression. This leads to confusing evaluation
error messages making it hard to accurately diagnose and then fix the
error.
For example, errors were reported as follows.
```
cannot coerce an integer to a string
1| let foo = 7; in "bar" + foo
| ^
cannot add a string to an integer
1| let foo = "bar"; in 4 + foo
| ^
cannot coerce an integer to a string
1| let foo = 7; in "x${foo}"
| ^
```
This commit changes the ExprConcatStrings expression vector to store a
sequence of expressions *and* their expansion locations so that error
locations can be reported accurately. For interpolation, the error is
reported at the beginning of the entire `${foo}`, not at the beginning
of `foo` because I thought this was slightly clearer. The previous
errors are now reported as:
```
cannot coerce an integer to a string
1| let foo = 7; in "bar" + foo
| ^
cannot add a string to an integer
1| let foo = "bar"; in 4 + foo
| ^
cannot coerce an integer to a string
1| let foo = 7; in "x${foo}"
| ^
```
The error is reported at this kind of precise location even for
multi-line indented strings.
This probably helps with at least some of the cases mentioned in #561
