4 ms·

Fmt – A Modern Formatting Library

44 points by int_hdlr 6y ago

billforsternz 6y ago
I know I am a barbarian, but I'll probably just continue to use my own std::string sprintf( const char *fmt, ... ); function as implemented here; https://github.com/billforsternz/retro-sargon/blob/master/src/util.cpp https://github.com/billforsternz/retro-sargon/blob/master/sr...
qppo 6y ago
(C++)
- throw0101a 6y ago
  I was wondering that was "modern" about it, given that fmt(1) has been around since 3BSD (1979?).
  - agallego 6y ago
    very different, in just about every way. Type safety, deduced by the compiler instead of the stringly-typed %d and friends for starters.
grandmczeb 6y ago
For those who don’t know, much of the fmt library interface has been standardized in C++20 as std::format. It’s a great library, and a very efficient alternative to ostream and friends.
- vitaut 6y ago
  In case anyone is interested here's brief history of standardizing std::format: http://www.zverovich.net/2019/07/23/std-format-cpp20.html http://www.zverovich.net/2019/07/23/std-format-cpp20.html
- mhh__ 6y ago
  It's great that it's there now but it absolutely amazes me how long some of these features have taken. Concepts took, what, the best part of 20 years and the end result is just OK. Unless there's something I'm missing the system as standardized in C++20 just seems overcooked - not very elegant while also introducing a lot of new legalese in the standard. The system D uses seems to be much simpler (and more powerful in the sense that it effectively moves template constraints into the library - doesn't require updating as traits change) i.e. template<typename T> if(constexprConditionOnT<T>) void toy(T x) {}
  - thechao 6y ago
    The real power of concepts was the “concept_map” construct. Gaby & Bjarne wanted predicated checking instead of structural checking. Without structural checking there’s no “natural” way in C++ syntax to specify a concept_map. The problem is that predicate checking isn’t that great. I can send you the turd of a PhD thesis that was left after my doctorate was bombed by the removal of concepts from C++0x.
nsajko 6y ago
I kind of think that IEEE 754 (or other binary floating point) to ASCII conversion should be its own project, or else in a language's standard library, since it is quite involved, especially if accuracy is important.
- zidel 6y ago
  The low level part of that was standardized in C++17 as std::to_chars https://en.cppreference.com/w/cpp/utility/to_chars https://en.cppreference.com/w/cpp/utility/to_chars
ncmncm 6y ago
I see that the ostream comparison is cooked. On each iteration they trigger a memory allocation.
dietrichepp 6y ago
It’s really astonishing how bad <ostream> is. It’s not just one or two problems—the <ostream> interface is just a total disaster. In my mind, fmt is not just a replacement for <ostream>, it’s a replacement for <cstdio>, which is a much higher bar to clear. https://www.moria.us/articles/iostream-is-hopelessly-broken/ https://www.moria.us/articles/iostream-is-hopelessly-broken/

44 points by int_hdlr 6y ago

I know I am a barbarian, but I'll probably just continue to use my own std::string sprintf( const char *fmt, ... ); function as implemented here; https://github.com/billforsternz/retro-sargon/blob/master/src/util.cpp https://github.com/billforsternz/retro-sargon/blob/master/sr...

qppo 6y ago

(C++)

throw0101a 6y ago
I was wondering that was "modern" about it, given that fmt(1) has been around since 3BSD (1979?).
- agallego 6y ago
  very different, in just about every way. Type safety, deduced by the compiler instead of the stringly-typed %d and friends for starters.

grandmczeb 6y ago

For those who don’t know, much of the fmt library interface has been standardized in C++20 as std::format. It’s a great library, and a very efficient alternative to ostream and friends.

vitaut 6y ago
In case anyone is interested here's brief history of standardizing std::format: http://www.zverovich.net/2019/07/23/std-format-cpp20.html http://www.zverovich.net/2019/07/23/std-format-cpp20.html
mhh__ 6y ago
It's great that it's there now but it absolutely amazes me how long some of these features have taken. Concepts took, what, the best part of 20 years and the end result is just OK. Unless there's something I'm missing the system as standardized in C++20 just seems overcooked - not very elegant while also introducing a lot of new legalese in the standard. The system D uses seems to be much simpler (and more powerful in the sense that it effectively moves template constraints into the library - doesn't require updating as traits change) i.e. template<typename T> if(constexprConditionOnT<T>) void toy(T x) {}
- thechao 6y ago
  The real power of concepts was the “concept_map” construct. Gaby & Bjarne wanted predicated checking instead of structural checking. Without structural checking there’s no “natural” way in C++ syntax to specify a concept_map. The problem is that predicate checking isn’t that great. I can send you the turd of a PhD thesis that was left after my doctorate was bombed by the removal of concepts from C++0x.

nsajko 6y ago

I kind of think that IEEE 754 (or other binary floating point) to ASCII conversion should be its own project, or else in a language's standard library, since it is quite involved, especially if accuracy is important.

zidel 6y ago
The low level part of that was standardized in C++17 as std::to_chars https://en.cppreference.com/w/cpp/utility/to_chars https://en.cppreference.com/w/cpp/utility/to_chars

ncmncm 6y ago

I see that the ostream comparison is cooked. On each iteration they trigger a memory allocation.

dietrichepp 6y ago

It’s really astonishing how bad <ostream> is. It’s not just one or two problems—the <ostream> interface is just a total disaster. In my mind, fmt is not just a replacement for <ostream>, it’s a replacement for <cstdio>, which is a much higher bar to clear. https://www.moria.us/articles/iostream-is-hopelessly-broken/ https://www.moria.us/articles/iostream-is-hopelessly-broken/