Make fields of DataBuf private

[kevinbackhouse]

This a refactoring change to tidy the DataBuf class. I made the pData_ and size_ fields private and added a few new methods which you can use to read and write the data.

[codecov]

Codecov Report

Merging #1886 (9ff72e5) into main (cb16324) will increase coverage by 0.06%.
The diff coverage is 64.08%.

@@            Coverage Diff             @@
##             main    #1886      +/-   ##
==========================================
+ Coverage   60.80%   60.86%   +0.06%     
==========================================
  Files          96       96              
  Lines       18963    19044      +81     
  Branches     9512     9726     +214     
==========================================
+ Hits        11531    11592      +61     
- Misses       5131     5139       +8     
- Partials     2301     2313      +12     

Impacted Files	Coverage Δ
src/convert.cpp	53.50% <0.00%> (ø)
src/exiv2.cpp	57.85% <0.00%> (ø)
src/mrwimage.cpp	41.89% <0.00%> (+0.55%)	⬆️
src/orfimage_int.cpp	33.33% <0.00%> (ø)
src/preview.cpp	50.35% <32.35%> (+1.27%)	⬆️
src/exif.cpp	65.90% <33.33%> (ø)
src/rafimage.cpp	20.00% <33.33%> (-0.74%)	⬇️
src/basicio.cpp	48.40% <40.00%> (-0.36%)	⬇️
src/tiffvisitor_int.cpp	76.23% <42.10%> (ø)
src/tags_int.cpp	76.34% <42.85%> (-0.13%)	⬇️
... and 24 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cb16324...9ff72e5. Read the comment docs.

[clanmills]

I think it's a good idea to change this as part of the new Exiv2/API for v1.00. Several suggestions. It's fine to ignore my suggestions!.

1. Provide a default offset=0 for c_str() and c_data().

const byte* c_data(size_t offset=0) const;

2. Is c_data() a good name? Perhaps data(offset=0) if easier on the eye.

3. It there merit in have having operator = for common assignments, such as:

unit64_t foo = dataBuf(size_t offset = 0,ByteOrder bo = inherit); 

4. There is a ByteOrder inherit. It's used by the binaryDecoder magic (and discussed in my book).

I'm wondering if we should have a ByteOrder on the DataBuf constructor.

On a read, the default bo=inherit would mean "use the buffer byteOrder used in the DataBuf constructor." We could avoid lots of ugly byteOrder arguments. For example, when you allocated an IFD, you store the byteOrder in the buffer itself and the code proceeds more smoothly. The byte swapping takes place in the buffer and not in the parser.

[kevinbackhouse]

@clanmills: Thanks, those are good suggestions.

• I'll add the size_t offset=0 default parameter and do a global search and replace on all the places where I have written .c_data(0). I'll do that in this PR.
• Do you have a suggestion for the name of the method? I added two methods named data() and c_data(). The latter returns a const pointer. (The vast majority of uses of pData_ are read-only.) Would cdata be a better name?
• I like the idea of DataBuf having a default byte order as a constructor parameter. I think it will be easier to add that in a separate follow-up PR though.

[piponazo]

Very nice contribution done here!

I particularly appreciate the following facts:

• Some reinterpret_cast went away.
• Calls to memset, memcpy, memcmp are now better encapsulated. (Maybe we can get rid of the inclusion of the cstring header in some places now).

As Robin mentioned it would be nice to try to improve the byteOrdering issue, but I agree that it would be better to handle it in a separate PR (this one is already quite big).

I made a comment about adding additional unit tests for the new code. But it can be also tackled in a separate PR. Actually, I could offer my help for that 😉

[piponazo]

In this test you are checking the "happy" cases. It would also be good to create additional tests to check what happens on the "non-happy" ones:

• Offset out of limits
• passing null pointers
• etc

[clanmills]

It's great to see this merged. Demoting endian handling to the buffer is a different project.