Cast fixed-length string -> std::string missing

[JATC1024]

Is your feature request related to a problem? Please describe.
H5Easy.load cannot load fixed length strings. I always get the error Can't output std::string as fixed-length. Use raw arrays or FixedLenStringArray. I tried to use HighFive::FixedLenStringArray<N> and it worked. However, I had to know the length of the strings beforehand since N must be a constant. This does not seem viable to me.

Describe the solution you'd like
Something like

int size = dataset.getStringSize();
HighFive::FixedLenStringArray strs(size);

would be nice. Or even better if H5Easy will handle everything so that I don't have to care if the dataset is fixed length or not.

[ferdonline]

Hey.
Your request is understandable but I believe HighFive already implements the required API.
Have you tried along the lines? (not tested):

int size = dataset.getDataType().getSize();
HighFive::FixedLenStringArray strs(size);

[JATC1024]

Thanks for replying. I think my comment here best describes my problem: #94 (comment).

Looking through the document, I cannot find a way to construct FixedLenStringArray without knowing N beforehand. Did I miss something?

[JATC1024]

I think the problem here is that FixedLenStringArray<N> is a std::vector<std::array<T, N>> in disguise, which requires knowing N in compile time.

Therefore,

const int len = 10;
HighFive::FixedLenStringArray<len> strs;

works but

const int len = dataset.getDataType().getSize();
HighFive::FixedLenStringArray<len> strs;

doesn't.

[ferdonline]

Oh, I'm sorry I overlooked the fact that the signature of FixedLenStringArray is FixedLenStringArray<N>, then N must be known at compile time.
I find your use case valid. There are plans to create a replacement of FixedLenStringArray but unfortunately our bandwidth is limited. If you feel comfortable to draft a solution, along the lines of using regular std::vector with std::string_view but keep compat with c++11 (like https://github.com/bitwizeshift/string_view-standalone) I would be happy to review.

[JATC1024]

I'll try to look into this in my free time. For now, I will simply set N as a big enough integer.

[tdegeus]

@JATC1024 You probably have a good reason to do what you do. Still, std::string may not be fixed-size, but is still pretty 'easy':

auto mystring = H5Easy::load<std::string>(file, "/path/to/my/string");

Also, unlike @ferdonline I'm not super sure that a 'fixed-size' string object that is dynamically allocated is relevant. Wouldn't that we std::string that you reserve to a certain size? Also this you can do currently (though not in H5Easy but in HighFive).
The current compile-time allocated truly fixed-size string is different. It presumably lives on the stack on might gain you some speed (though I doubt a lot, because I think that disk-IO will always be limiting).

[JATC1024]

@tdegeus
Oh I think I didn't describe the problem properly. I want to load an array of strings, not just a single string. What I'm trying to do is:

std::vector<std::string> strs;
strs = H5Easy::load<std::vector<std::string>>(file, "/path/to/strings")`

For variable-length strings, this works. For fixed-length strings, it shows the error Can't output std::string as fixed-length. Use raw arrays or FixedLenStringArray.

Or are you suggesting that I can load each string in the array individually? Something like

auto mystring = H5Easy::load<std::string>(file, "/path/to/strings", index);

I see you do have a function like that. I will try it tomorrow, though I think I will get the same error.

[tdegeus]

Thanks for the clarification @JATC1024 . This makes a lot of sense. I should say that I'm not at all an expert on HDF5 fixed-length strings. That being said, I see no reason why HighFive should not provide a method to load strings that are stored as fixed-length to std::string. I think that this should be implemented in the 'core' for std::string types. In that case it should be automatically available in H5Easy as well.

[tdegeus]

Shall we rename this issue (or open a new one) indicating that it is more than missing conversion from fixed-length strings to std::string?

[JATC1024]

@tdegeus sure, whatever way you see fit. If you create a new one then I think we can close this one.

[1uc]

With #744 merged std::strings and containers thereof, e.g. std::vectorstd::string can be read and written to fixed or variable length HDF5 strings. There's an example here:
https://github.com/BlueBrain/HighFive/blob/master/src/examples/read_write_std_strings.cpp