Change &foo[0] to foo.data() for a std::vector

[eric-hughes-tiledb]

The idiom &foo[0] is fine for an array; it's not the same for a vector. Accessing a non-existent element through operator[] is undefined behavior, triggered when foo is empty. It's compliant behavior for a library to throw an assertion failure in this case (that's how the defect came to light). The standard library provides data() to substitute for the old, C-style idiom.

---

TYPE: BUG
DESC: Treating std::vector like an array; accessing an element that's not present to get its address.

[joe-maley]

LGTM at face value. If this was actually triggering an error, we should put a check that data is non-empty.

[joe-maley]

Also note that you'll need to clang-format this, but I assume you will see that in the related CI failures.

[dhoke4tdb]

LGTM at face value. If this was actually triggering an error, we should put a check that data is non-empty.

My recollection is that the called set_buffer() will properly handle empty .data() (being also passed data.size()), that the error is in appropriate use of vector methods, not a problem with the actual tiledb data.

[eric-hughes-tiledb]

the error is in appropriate use of vector methods

That's exactly right. The empty-buffer case was handled in #590.