Sometimes people don't remember the exact number of decimal places
supported by denomination or don't want to count (e.g. when converting
fiat to BTC the calculator may yield too precise value). It's helpful to
say in error message at which digit the precision is too high.
This adds `TooPreciseError` struct containing the information and
improves the error message to display it.
This variant lacked pretty important information: what is the limit. We
could just write it in the error message but it's even better to move it
to a separate struct which also says how many characters exceed the
limit - this helps users know how many need to be deleted.
The `InvalidFormat` variant was pretty bad: it didn't make it clear what
exactly is wrong with the input string, especially since it was used
when the denomination was missing. Not only was this unhelpful to the
users who don't know they have to write the denomination when the amount
was otherwise correct but it was also attributed to a problem with the
amount rather than a problem with denomination.
To fix this the variant is replaced by `MissingDigitsError`,
`MissingError` and `InvalidCharError` - all the cases `InvalidFormat`
was originally used in.
`InvalidCharError` is effectively the same as the existing variant but
it was made into a separate error to enable special casing `.` and make
extensions possible. Further this opportunity was used to add a special
case for `-` as well.
`MissingDigitsError` currently contains special casing for empty string
and a string only containing minus sign. It's currently unclear if it's
useful so this change makes this distinction private and only makes it
affect error messages.
As opposed to the previous two variants, `MissingDenominationError` was
added to `ParseError`. The struct is currenly empty and may be extended
in the future with e.g. span information.
df1d2f6eb5 Add unchecked variants to Amount and SignedAmount (yancy)
Pull request description:
The checked variants have worse performance than the unchecked variants due to the additional branching operations. To improve performance where overflow is either not possible or not a concern, unchecked variants of `Amount` and `SignedAmount` are introduced for addition, subtraction and multiplication.
Note, it seems the default behavior for the test framework is to panic on overflow, so I haven't figured out a good way to add tests for this. Marking as a draft for now.
closes: https://github.com/rust-bitcoin/rust-bitcoin/issues/2434
ACKs for top commit:
Kixunil:
ACK df1d2f6eb5
apoelstra:
ACK df1d2f6eb5 gonna go ahead and merge this, we can revisit if necessary when we look at `units` overflow behavior in general
Tree-SHA512: 3fbb0ec81a758b350569226c44e25f6ca49e551566bee83c05c1c2b343874ef657d63a36b5f51c41582d8a8e36466275c574ebff6d363ed7c112ac8b4d5376fa
The checked variants have worse performance than the unchecked variants
due to the additional branching operations. To improve performance where
overflow is either not possible or not a concern, unchecked variants
of Amount and SignedAmount are introduced.
251579f4ef feat: implement TryFrom trait to SignedAmount and Amount (Sumit Kumar)
Pull request description:
Closes: #2245
Adds `TryFrom<SignedAmount> for Amount` and `TryFrom<Amount> for Amount` in units module
ACKs for top commit:
tcharding:
ACK 251579f4ef
Kixunil:
ACK 251579f4ef
apoelstra:
ACK 251579f4ef
Tree-SHA512: 3e58d7a891019ccd272417eadc977037167439e3385a7b47c060fe93eba9c47fc37cdc0ca5551174ef2d93b256b2804ad7c01f5f2470ef9e9b7b912877aed11c
c69caafefc Remove attribute comments (Tobin C. Harding)
3e83ef9276 Remove consensus error wrapper TODO (Tobin C. Harding)
bfabea94e9 Remove unwrap comment (Tobin C. Harding)
8bdaf4a34d Remove carrying_mul TODO (Tobin C. Harding)
Pull request description:
Add issues and remove the TODOs from the code.
Resolves: #2368
ACKs for top commit:
apoelstra:
ACK c69caafefc
Kixunil:
ACK c69caafefc
Tree-SHA512: b10a3de8da7ace890735023f8441605dd11b0227c27a2357556b8aaa8276a7f34ed220e3bcbc93aad4b35357319318ff7de27210e8f60dd90f6c55af23e21470
0997382772 io: Enable alloc from std (Tobin C. Harding)
ba1166a63b Make crate level attributes uniform (Tobin C. Harding)
Pull request description:
Make the trait level attributes uniform across all released crates in the repo. Excludes things that are obviously not needed, eg, bench stuff if there is not bench code.
- Remove `uninhabited_references` - this is allow by default now.
- Remove `unconditional_recursion` and mark the single false positive we have with an `allow`.
Note, this does not add `missing_docs` to the `io` crate. There is an open PR at the moment to add that along with the required docs.
ACKs for top commit:
apoelstra:
ACK 0997382772
Kixunil:
ACK 0997382772
Tree-SHA512: ef1f638aca171536287cce369be98998e871d26468ad2d8c39d9004db610b406471809c283540a4a19bcede78b12b8976a1bb37e5d431fbff8c8a3e53a64d4e3
Recently (in #2379) we patched the `ParseAmountError` but we don't check
kani code on every pull request so we broke it.
Fix kani test to use the new `OutOfRangeError`.
Close: #2424
Make the trait level attributes uniform across all released crates in
the repo. Excludes things that are obviously not needed, eg, bench stuff
if there is not bench code.
- Remove `uninhabited_references` - this is allow by default now.
- Remove `unconditional_recursion` and mark the single false positive we
have with an `allow`.
Note, this does not add `missing_docs` to the `io` crate. There is an
open PR at the moment to add that along with the required docs.
Previously the crate used negative reasoning to enable `std` which was
hard to understand, required the `prelude` module and wasn't really
needed because it's only needed when a crate wants to add `alloc`
feature-backwards compatibly and this crate always had the feature.
This cleans up usage to unconditionally use `#[no_std]` and then just
add `extern crate` on top as needed by activated features.
Using the crate without allocation was previously disabled making the
crate empty without the feature. This chage makes it more fine-grained:
it only disables string and float conversions which use allocator. We
could later provide float conversions by using a sufficiently-long
`ArrayString`.
7bf478373a Model `TooBig` and `Negative` as `OutOfRange` (Martin Habovstiak)
54cbbf804f Express `i64::MAX + 1` as `i64::MIN.unsigned_abs()` (Martin Habovstiak)
b562a18914 Move denomination error out of `ParseAmountError` (Martin Habovstiak)
5e6c65bc1a Clean up `unsigned_abs` (Martin Habovstiak)
Pull request description:
Closes#2265Closes#2266
Disclaimer: I did this in December and don't remember why I haven't pushed it. Maybe because it's somehow broken but I don't see how so please review a bit more carefully just in case.
ACKs for top commit:
tcharding:
ACK 7bf478373a
apoelstra:
ACK 7bf478373a
Tree-SHA512: 1f6e9adae9168bd045c9b09f06d9a69efd47ccc7709ac9ecaf48cb86e265b448b9b52a199ac5e6838d5029f5bc7514c5d7deb15a4d7c8a4606a353f390745570
The error returned when parsing amount had a `Negative` variant which
was weird/unreachable when parsing `SignedAmount`. Also weirdly, parsing
would return `TooBig` when the amount was negative - too low.
To resolve this we merge them into one `OutOfRange` variant that nuges
the consumers to make principled decisions and print error messages as
amounts being more than or less than a specific value which is easier to
understand for the users. Notably, the API still allows getting
information about which type was parsed and which bound was crossed but
in a less obvious way. This is OK since users who have a very good
reason can use this information but most won't.
Closes#2266
The `from_str_in` methods on amounts returned `ParseAmountError` which
contained `InvalidDenomination` among its variants. This one was never
returned because the method doesn't parse denomination.
This change separates the error out.
Closes#2265
Previousle we copied `unsigned_abs` method from `core` because it was
unstable in older MSRV. Our current MSRV allows using the method
directly so this removes our old one and uses the one from standard
library instead.
b2344e019d units: Assert roundtrip SignedAmount/str overflows (Tobin C. Harding)
baadcf4c0a units: Test that SignedAmount float conversion overflows (Tobin C. Harding)
d768f25da8 units: Remove duplicate assertion (Tobin C. Harding)
1d536ac8b2 units: Enable parsing Amount from u64::MAX (Tobin C. Harding)
Pull request description:
Our `Amount` type uses an internal `u64` and we maintain no invariants on the inner value. Therefore we should be able to parse `u64::MAX`.
Fix the parsing code by removing the explicit, incorrect check and fix unit tests to mirror this behaviour.
Fix: #2297
ACKs for top commit:
Kixunil:
ACK b2344e019d
apoelstra:
ACK b2344e019d
Tree-SHA512: 944f8d0bfedc559f0444f75eca7d3fba042fbc204c4c032d09ff0edc29be280a3707f5b363dbc04f0d7bdf64701c0c4619e2e0de683d804a2663c2a20ac963f6
We should not be able to roundtrip a `SignedAmount` value greater than
`MAX`, add a test to prove so.
While we are at it document the assertion above that proves we can parse
a float representing an `Amount` greater than `SignedAmount::MAX`.
`Denomination::Bitcoin` and `Denomination::Satoshi` are often used,
especially in test code so this change adds `BTC` and `SAT` - short,
readable constants. Notably this doesn't add the other constants as that
would lead to either unidiomatic names or confusing casing (MSAT meaning
millisat not megasat) and they are not used that much anyway.
Our `Amount` type uses an internal `u64` and we maintain no invariants
on the inner value. Therefore we should be able to parse `u64::MAX`.
Fix the parsing code by removing the explicit, incorrect check and fix
unit tests to mirror this behaviour.
Fix: #2297
f06d12455f bitcoin: Remove the custom sink (Tobin C. Harding)
b503aa1544 Run the formatter (Tobin C. Harding)
3ca55fb163 Remove qualifying path from Read and Write (Tobin C. Harding)
ebeb21fa7a Import fmt::Write using underscore (Tobin C. Harding)
e2dbcb1d28 Use W for writer generic type (Tobin C. Harding)
8704d9f0ae docs: Fix grammar (Tobin C. Harding)
Pull request description:
A few cleanups to how we use the `io` crate, this is reasonably trivial but commit `a6c7e696 Remove qualifying path from Read and Write` is big, I have however gone to some effort to make sure it is easy to flick through the diff.
Done in preparation for another go at the `BufRead` stuff.
ACKs for top commit:
apoelstra:
ACK f06d12455f
Kixunil:
ACK f06d12455f
Tree-SHA512: 751c489c67901c7563f1cc91f7761a4e3c276ae1981010338134e8c13200720ba69fcc74948c1dc1e6e65390197da0da27b2b69b86034029748321b404142cba
When we use the `fmt::Write` trait it is just to call its methods, we
can therefore, without any change to the logic, use `as _` when
importing the trait. This prevents naming conflicts.
Done in preparation for importing the `io::Write` trait.
Martin Habovstiak
Tobin C. Harding
Andrew Poelstra
yancy
Sumit Kumar
Fmt Bot
Jiri Jakes