Especially when dealing with big files the performance is quite bad. This
change speeds it up a little by using an unordered map to find the elements
for certain offsets instead of using linear lookup.
According to callgrind the modified functions where one with the biggest
own cost. The function updateRelativeOffsets() was actually the 2nd costly
function just below __memcpy_avx_unaligned_erms.
Now TagParser::EbmlElement::internalParse(TagParser::Diagnostics&) is quite
high as well as standard IO stream functions.
Different media/tag formats specify languages and countries
differently. This change introduces a Locale class to keep track
of the format being used. So far there are no automatic conversions
implemented so it is entirely up to the user to pass valid values using
a format which matches the one required by the media/tag format.
This change also adds support for Matroska's IETF elements so at least the
raw value can be read, written and is preserved.
For now let's just ignore these elements explicitely until they are
actually supported. This way the warnings are at least more specific and
there will be no inconsistency when updating the track language.
1. Convert TYER and related fields of old ID3v2 versions to the new TDRC
field and only expose that via the generic accessors.
2. When writing an old ID3v2 tag, convert TDRC back to the old fields.
3. One can still manually unset the via 1. auto-populated TDRC to disable 2.
and write the old fields directly. So the automatic handling does not
reduce the flexibility of the library.
4. Deprecate 'Year'; it is replaced by the already existing 'RecordDate'
which is now supposed to be used everywhere where 'Year' was used before
5. Introduce 'ReleaseDate' to support this field which is supported in
ID3v2.4.0 and Matroska via the generic accessors.
6. Use ISO format when converting tag values of the type DateTime to/from
string. This is closer to what's used in ID3v2 tags internally. (The
library still allows the old format as fallback when parsing for
compatibility.)
* Close or flush streams explicitely so writing is not
deferred
* to catch errors in the right place
* to avoid suppressing errors completely when writing
would be deferred to the destructor invocation
* Improve comments
This should fix all non-erros, leaving only warnings which
are indeed potential problems.
The following warnings should be safe to ignore:
* Conversions of various offsets from uint64 to
std::streamoff/int64 are safe because such offsets have
been obtained via tellg() and other functions
returning std::streamoff in the first place.
* It also works vice-versa since tellg() should not
return negative offsets with exceptions enabled.
* Conversions from char to unsigned char are also ok.
* Unused diag arguments can be ignored (those might be
useful later).
* Annotate all intended fallthoughs.
Seems like the trick with the three dots isn't working
with GCC 8 anymore. So let's make it a default template
parameter then.
Not sure whether GCC 8 is here correct and whether this
workaround causes further trouble.
Martchus