More module tests - #2309
More module tests#2309
Conversation
1a2b2d1 to
d5203d3
Compare
vitaut
left a comment
vitaut
left a comment
There was a problem hiding this comment.
Thanks for the PR. Having a module test seems like a good idea but I am concerned about duplication. I think we don't need to replicate so many test cases here, it's enough to exercise every user-facing public function like format or format_to once. In particular, we shouldn't be testing multiple built-in argument types because they all go through the same path.
Also there are some json files that I don't think should be here.
|
Right, these two additional files sneaked in by accident. Thanks for pointing this out. Regarding duplication: it's not just the name like e.g. If you happen to have sections to point me to which concern you I will look into them and remove superfluous duplicates. I'd love to have a tool to aide me but I'm not aware of one. |
| EXPECT_EQ("42", fmt::format("{}", 42)); | ||
| EXPECT_EQ("42", fmt::format("{:}", 42u)); | ||
| EXPECT_EQ("+42", fmt::format("{:+d}", short{42})); | ||
| EXPECT_EQ("42", fmt::format("{:d}", unsigned short{42u})); | ||
| EXPECT_EQ("-42", fmt::format("{0}", -42L)); | ||
| EXPECT_EQ("2a", fmt::format("{0:x}", 42uL)); | ||
| EXPECT_EQ("***42", fmt::format("{:*>5}", 42LL)); | ||
| EXPECT_EQ("0X2A", fmt::format("{:#X}", 42uLL)); | ||
| EXPECT_EQ("===42.000000===", fmt::format("{:=^15f}", 42.0)); | ||
| EXPECT_EQ("00000042.1", fmt::format("{:<010.5G}", 42.1f)); | ||
| EXPECT_EQ(" +4.21235e+11", fmt::format("{:+{}.{}e}", 42.1234567e10L, 13, 5)); | ||
| EXPECT_EQ("-0x1.5000000000000p+5", fmt::format("{: a}", -42.0)); | ||
| constexpr auto nan = std::numeric_limits<double>::quiet_NaN(); | ||
| EXPECT_EQ("-nan-", fmt::format("{:-^5}", nan)); | ||
| constexpr auto inf = std::numeric_limits<double>::infinity(); | ||
| EXPECT_EQ("-INF-", fmt::format("{:-^5F}", inf)); | ||
| EXPECT_EQ("-INF ", fmt::format("{: 5G}", -inf)); | ||
| EXPECT_EQ("42", fmt::format("{:s}", "42")); | ||
| EXPECT_EQ("a", fmt::format("{:c}", 'a')); | ||
| EXPECT_EQ("a", fmt::format("{:c}", unsigned char{'a'})); | ||
| EXPECT_EQ("a", fmt::format("{:c}", signed char{'a'})); | ||
| EXPECT_EQ("false", fmt::format("{}", false)); | ||
| EXPECT_EQ("0B1", fmt::format("{:#B}", true)); | ||
| EXPECT_EQ("0x2a", fmt::format("{:p}", (void *)42)); | ||
| EXPECT_EQ("1234", fmt::format("{:L}", 1234)); |
There was a problem hiding this comment.
For example here we don't need to test all argument types. Since they all go through the same code path, it is very unlikely that one of them works and another doesn't. I suggest picking one type (e.g. int) and testing public methods with it. In the unlikely case that we find some problem in the future, we can add a regression test.
d5203d3 to
700551f
Compare
|
Your suggestions are in. This is also updated to msvc 16.11-pre1 |
700551f to
d8e200d
Compare
vitaut
left a comment
vitaut
left a comment
There was a problem hiding this comment.
Your suggestions are in. This is also updated to msvc 16.11-pre1
Thank you. A few more comments, otherwise looks great.
| EXPECT_FALSE(no_args.get(1)); | ||
|
|
||
| using ctx = fmt::format_context; | ||
| fmt::basic_format_args<ctx> args = fmt::make_format_args<ctx>(42); |
There was a problem hiding this comment.
ctx can be dropped here because it is the default. Same in multiple places below.
| } | ||
|
|
||
| TEST(module_test, ptr) { | ||
| auto p = (int*)42; |
There was a problem hiding this comment.
Please use more narrow bit_cast or reinterpret_cast instead of a C-style cast.
| } | ||
|
|
||
| TEST(module_test, weekday) { | ||
| const auto de = std::locale("de_DE"); |
There was a problem hiding this comment.
This may throw if the locale is not available.
There was a problem hiding this comment.
Yeah ...
Luckily we don't care here.
There was a problem hiding this comment.
Why don't we care? Since this is a non-functional test I suggest using the std::locale::classic() which is always available.
There was a problem hiding this comment.
We don't care for the same reason why you suggest using the classic locale: it just has to look up the correct overloads. If you prefer std::locale::classic() then I'll change it accordingly.
There was a problem hiding this comment.
it just has to look up the correct overloads.
It would enough if we didn't actually run this code, only compile. But my understanding is that we eventually want to run this code.
If you prefer std::locale::classic() then I'll change it accordingly.
Please do.
| static_assert(fmt::is_formattable<disabled_formatter_convertible>::value, ""); | ||
| } | ||
|
|
||
| struct convertible_to_int { |
There was a problem hiding this comment.
Everything starting from here to the end of the file falls into the category of "functional testing", please remove. It is very unlikely that this functionality behaves differently between modular and non-modular build because it goes through the same argument handling paths as everything else.
There was a problem hiding this comment.
I've condensed it as much as possible to name-lookup.
d8e200d to
865994b
Compare
core.h, format.h, args.h, chrono.h, color.h, printf.h, os.h
865994b to
576f0b7
Compare
|
Thank you! |
2 participants
Most of the API of core.h, format.h, and args.h
Just narrow strings. Tests for wide strings will come later.