# Description
Our `ShellError` at the moment has a `std::mem::size_of<ShellError>` of
136 bytes (on AMD64). As a result `Value` directly storing the struct
also required 136 bytes (thanks to alignment requirements).
This change stores the `Value::Error` `ShellError` on the heap.
Pro:
- Value now needs just 80 bytes
- Should be 1 cacheline less (still at least 2 cachelines)
Con:
- More small heap allocations when dealing with `Value::Error`
- More heap fragmentation
- Potential for additional required memcopies
# Further code changes
Includes a small refactor of `try` due to a type mismatch in its large
match.
# User-Facing Changes
None for regular users.
Plugin authors may have to update their matches on `Value` if they use
`nu-protocol`
Needs benchmarking to see if there is a benefit in real world workloads.
**Update** small improvements in runtime for workloads with high volume
of values. Significant reduction in maximum resident set size, when many
values are held in memory.
# Tests + Formatting
# Description
* I was dismayed to discover recently that UnsupportedInput and
TypeMismatch are used *extremely* inconsistently across the codebase.
UnsupportedInput is sometimes used for input type-checks (as per the
name!!), but *also* used for argument type-checks. TypeMismatch is also
used for both.
I thus devised the following standard: input type-checking *only* uses
UnsupportedInput, and argument type-checking *only* uses TypeMismatch.
Moreover, to differentiate them, UnsupportedInput now has *two* error
arrows (spans), one pointing at the command and the other at the input
origin, while TypeMismatch only has the one (because the command should
always be nearby)
* In order to apply that standard, a very large number of
UnsupportedInput uses were changed so that the input's span could be
retrieved and delivered to it.
* Additionally, I noticed many places where **errors are not propagated
correctly**: there are lots of `match` sites which take a Value::Error,
then throw it away and replace it with a new Value::Error with
less/misleading information (such as reporting the error as an
"incorrect type"). I believe that the earliest errors are the most
important, and should always be propagated where possible.
* Also, to standardise one broad subset of UnsupportedInput error
messages, who all used slightly different wordings of "expected
`<type>`, got `<type>`", I created OnlySupportsThisInputType as a
variant of it.
* Finally, a bunch of error sites that had "repeated spans" - i.e. where
an error expected two spans, but `call.head` was given for both - were
fixed to use different spans.
# Example
BEFORE
```
〉20b | str starts-with 'a'
Error: nu:🐚:unsupported_input (link)
× Unsupported input
╭─[entry #31:1:1]
1 │ 20b | str starts-with 'a'
· ┬
· ╰── Input's type is filesize. This command only works with strings.
╰────
〉'a' | math cos
Error: nu:🐚:unsupported_input (link)
× Unsupported input
╭─[entry #33:1:1]
1 │ 'a' | math cos
· ─┬─
· ╰── Only numerical values are supported, input type: String
╰────
〉0x[12] | encode utf8
Error: nu:🐚:unsupported_input (link)
× Unsupported input
╭─[entry #38:1:1]
1 │ 0x[12] | encode utf8
· ───┬──
· ╰── non-string input
╰────
```
AFTER
```
〉20b | str starts-with 'a'
Error: nu:🐚:pipeline_mismatch (link)
× Pipeline mismatch.
╭─[entry #1:1:1]
1 │ 20b | str starts-with 'a'
· ┬ ───────┬───────
· │ ╰── only string input data is supported
· ╰── input type: filesize
╰────
〉'a' | math cos
Error: nu:🐚:pipeline_mismatch (link)
× Pipeline mismatch.
╭─[entry #2:1:1]
1 │ 'a' | math cos
· ─┬─ ────┬───
· │ ╰── only numeric input data is supported
· ╰── input type: string
╰────
〉0x[12] | encode utf8
Error: nu:🐚:pipeline_mismatch (link)
× Pipeline mismatch.
╭─[entry #3:1:1]
1 │ 0x[12] | encode utf8
· ───┬── ───┬──
· │ ╰── only string input data is supported
· ╰── input type: binary
╰────
```
# User-Facing Changes
Various error messages suddenly make more sense (i.e. have two arrows
instead of one).
# Tests + Formatting
Don't forget to add tests that cover your changes.
Make sure you've run and fixed any issues with these commands:
- `cargo fmt --all -- --check` to check standard code formatting (`cargo
fmt --all` applies these changes)
- `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A
clippy::needless_collect` to check that you're using the standard code
style
- `cargo test --workspace` to check that all tests pass
# After Submitting
If your PR had any user-facing changes, update [the
documentation](https://github.com/nushell/nushell.github.io) after the
PR is merged, if necessary. This will help us keep the docs up to date.
# Description
While perusing Value.rs, I noticed the `Value::int()`, `Value::float()`,
`Value::boolean()` and `Value::string()` constructors, which seem
designed to make it easier to construct various Values, but which aren't
used often at all in the codebase. So, using a few find-replaces
regexes, I increased their usage. This reduces overall LOC because
structures like this:
```
Value::Int {
val: a,
span: head
}
```
are changed into
```
Value::int(a, head)
```
and are respected as such by the project's formatter.
There are little readability concerns because the second argument to all
of these is `span`, and it's almost always extremely obvious which is
the span at every callsite.
# User-Facing Changes
None.
# Tests + Formatting
Don't forget to add tests that cover your changes.
Make sure you've run and fixed any issues with these commands:
- `cargo fmt --all -- --check` to check standard code formatting (`cargo
fmt --all` applies these changes)
- `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A
clippy::needless_collect` to check that you're using the standard code
style
- `cargo test --workspace` to check that all tests pass
# After Submitting
If your PR had any user-facing changes, update [the
documentation](https://github.com/nushell/nushell.github.io) after the
PR is merged, if necessary. This will help us keep the docs up to date.
* Remove unnecessary `#[allow]` annots
Reduce the number of lint exceptions that are not necessary with the
current state of the code (or more recent toolchain)
* Remove dead code from `FileStructure` in nu-command
* Replace `allow(unused)` with relevant feature switch
* Deal with `needless_collect` with annotations
* Change hack for needless_collect in `from json`
This change obviates the need for `allow(needless_collect)`
Removes a pessimistic allocation for empty strings, but increases
allocation size to `Value`
Probably not really worth it.
* Revert "Deal with `needless_collect` with annotations"
This reverts commit 05aca98445.
The previous state seems to better from a performance perspective as a
`Vec<String>` is lighter weight than `Vec<Value>`
