Floating-point formatting (#360) #866
Conversation
This is an attempt to fix #360. The algorithm produces decimal representations which are guaranteed to roundtrip and in ~99.8% actually produces the shortest possible representation. So this is a nice compromise between using a precision of digits10 and max_digits10. Note 1: The implementation only works for IEEE single/double precision numbers. So the old implementation is kept for compatibility with non-IEEE implementations and 'long double'. Note 2: If number_float_t is 'float', not all serialized numbers can be recovered using strtod (strtof works, though). (There is exactly one such number and the result is off by 1 ulp.) This can be avoided by changing the implementation (the fix is trivial), but then the resulting decimal numbers are not exactly short. Note 3: The code should be replaced by std::to_chars once it is available.
|
What is the license of the included conversion code? |
|
The conversion code is based on the reference implementation by Florian Loitsch which is distributed under the MIT license. I added a NOTICE file containing the license but obviously missed to commit it. Is it ok to add a NOTICE file? Or add the license to then end of json.hpp? |
|
Thanks a lot for the PR. I am currently a bit torn about merging it for the 3.0.0 release, because it involves quite a lot code and I won't have time for a thorough review until then. Furthermore, it could benefit from #700 as we can move this code in a separate (development) header. On the other hand, it does change the behavior of the library (albeit for the better), so maybe that adding it after 3.0.0 would mean to quickly release a 4.0.0 version. Any thoughts on this? |
nlohmann
added
the
state: please discuss
|
I would say that this is a bug fix not affecting the public API, and thus fixing it in a point release would be fine under SemVer guidelines. |
|
I would even say that that using |
|
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. |
stale
bot
added
the
state: stale
nlohmann
removed
the
state: stale
|
Is there still some interest in this PR? I'm asking because I tried to rebase this PR and failed miserably. Two times... :-) So if there is any interest, I'd like to close this and open a new PR (which seems a lot easier for me...) |
|
Yes, please! Now that the header is split, we can focus on adding such features. |
|
Ok. Where shall I put the implementation? I thought about something like a |
|
Sounds good. |
Try to fix #360.
printfdoesn't have a "to shortest" mode andstd::to_charsis not available yet. Use the Grisu2 algorithm for floating-point formatting which is a nice compromise between usingprintfwith a precision ofdigits10andmax_digits10, i.e. the resulting decimal representation is guaranteed to round-trip and is almost always the shortest such representation.This PR is mostly to see if everything compiles fine and check that all tests are passing. The change is quite large. But maybe it serves as a starting point for an alternative to always use
max_digits10.