Re-template json_pointer on string type

[falbrechtskirchinger]

This PR changes the json_pointer template parameter from basic_json to string type and deals with unexpected complications.

Most notably, a helper function for string concatenation is introduced to allow for some flexibility in the implementation of string types (and eliminate some temporaries as a positive side-effect). I'll leave it to others to remove more std::string API dependence (some low-hanging fruits are empty(), find_first_of(), substr()).

To Do:

• Remove context parameter default value as per greg's suggestion in the previous PR.
• Add backwards compatibility with json_pointer<basic_json>.
• Update documentation.
• Fix issue references in commits once ready for merge. (Don't want to spam GH issues as much.)

[coveralls]

Coverage remained the same at 100.0% when pulling fdb15f9 on falbrechtskirchinger:refactor-json_ptr-tpl into 261cc4e on nlohmann:develop.

[falbrechtskirchinger]

@nlohmann Can you please take a look at the Codacy report and tell me which tool is generating that strlen warning? This link is not accessible to me: https://app.codacy.com/gh/nlohmann/json/patterns/list?selectedEngine=81806a42-1d70-40e6-ad07-9a1a9da9e500#pattern-6542

[nlohmann]

@nlohmann Can you please take a look at the Codacy report and tell me which tool is generating that strlen warning? This link is not accessible to me: https://app.codacy.com/gh/nlohmann/json/patterns/list?selectedEngine=81806a42-1d70-40e6-ad07-9a1a9da9e500#pattern-6542

Sure:

Looks like a false positive - I ignored the issue. This should become effective in the next check.

[falbrechtskirchinger]

@nlohmann Can you please take a look at the Codacy report and tell me which tool is generating that strlen warning? This link is not accessible to me: https://app.codacy.com/gh/nlohmann/json/patterns/list?selectedEngine=81806a42-1d70-40e6-ad07-9a1a9da9e500#pattern-6542

(...)
Looks like a false positive - I ignored the issue. This should become effective in the next check.

Okay, thanks, I would have added // Flawfinder: ignore now that I know the tool.

[gregmarr]

Just a thought, instead of const char *cstr, could this be template<int N> ... const char (&cstr)[N] so that we avoid the strlen altogether?

[falbrechtskirchinger]

Just a thought, instead of const char *cstr, could this be template<int N> ... const char (&cstr)[N] so that we avoid the strlen altogether?

We'd still need const char * for things like type_name() which returns const char * (*) and then one'd have to handle the ambiguity between the two, either by introducing a helper type with template specialization for both cases or using enable_if/SFINAE.
I considered it but it's not worth the hassle in my opinion.

(*) I mean, we don't have to but it's nice to avoid std::string(type_name()).

[falbrechtskirchinger]

Once this passes review, prior to merge, I'd just like to do one last force-push to update commit messages with the issue references.

Edit: Forgot to include json_pointer/string_t.md. Fixed.

[gregmarr]

We'd still need const char * for things like type_name()

Agreed that it only makes sense if the only things are string literals and string objects, but doesn't make sense if there are true raw pointers.

[falbrechtskirchinger]

I discovered a few minor things in my full review. Will fix momentarily.

[falbrechtskirchinger]

Ready for final review and merge.

[falbrechtskirchinger]

@nlohmann 3.10.6 is now 3.11.0, right? So I need to update the version numbers?

[nlohmann]

@nlohmann 3.10.6 is now 3.11.0, right? So I need to update the version numbers?

Yes, so much has piled up. But I would have gone throw the release anyway to clean this up.

[falbrechtskirchinger]

You probably missed my comment on this closed PR: #3418 (comment)
Interested?

[nlohmann]

You probably missed my comment on this closed PR: #3418 (comment)
Interested?

Not right now.

[falbrechtskirchinger]

Version numbers updated. Mixed in some accidental edits that were supposed to go into a different branch and fixed that too.

[nlohmann]

Looks good to me.

[nlohmann]

Thanks!