Add codec for tag 0xC4A5 PrintImageMatching #81
Conversation
As the information about this tag is pretty obscure in the internet the codec is fairly basic, just checking the header, parsing the version and keeping the raw bytes.
georgethebeatle
force-pushed
the
pr-print-image-matching
branch
from
28996fb
to
e7f5402
Compare
| @@ -9,10 +9,7 @@ go 1.12 | |||
| require ( | |||
| @@ -1,11 +1,17 @@ | |||
| github.com/dsoprea/go-exif/v2 v2.0.0-20200321225314-640175a69fe4/go.mod h1:Lm2lMM2zx8p4a34ZemkaUV95AnMl4ZvLbCUbwOvLC2E= | |||
| log.Panicf("invalid header for tag 0xC4A5 PrintImageMatching") | ||
| } | ||
|
|
||
| versionLen := bytes.IndexByte(rawBytes[8:], 0) |
There was a problem hiding this comment.
Validate that versionLen is not -1.
|
|
||
| valueContext.SetUndefinedValueType(exifcommon.TypeByte) | ||
| rawBytes, err := valueContext.ReadBytes() | ||
| ev.Value = rawBytes |
There was a problem hiding this comment.
The documentation describes how to find an entry count and the list of entries. We can parse things better than just yielding an opaque byte slice, yes?
There was a problem hiding this comment.
I could not find any decent documentation on PrintIM. It is not specified in the exif 2.2 spec. The only thing I found was the link I posted above and it does not agree with my images. My images have a 106 byte PrintIM. According to the spec this would mean
Header: 8
Version: 5
ExtraNull: 1
EntryCount: 2
Entries: 6*EntryCount
This works out if there are 15 entries. However here is the actual PrintIM bytes of one of my photos:
00000000: 5072 696e 7449 4d00 3033 3030 0000 0300 PrintIM.0300....
00000010: 0200 0100 0000 0300 2200 0000 0101 0000 ........".......
00000020: 0000 0911 0000 1027 0000 0b0f 0000 1027 .......'.......'
00000030: 0000 9705 0000 1027 0000 b008 0000 1027 .......'.......'
00000040: 0000 011c 0000 1027 0000 5e02 0000 1027 .......'..^....'
00000050: 0000 8b00 0000 1027 0000 cb03 0000 1027 .......'.......'
00000060: 0000 e51b 0000 1027 0000 0a .......'...
Bytes 15-16 are [03 00] which yields 3 entries (order is Little Endian). So it does not work out.
Let's say we ignore the count and treat the remaining bytes as 6 byte entries. Here is what we get:
0200 0100 0000
0300 2200 0000
0101 0000 0000
0911 0000 1027
0000 0b0f 0000
1027 0000 9705
0000 1027 0000
and so on.
If the first 2 bytes of each entry would designate its tag id we have 2 tag ids of [00 00] which does not make much sense as we have a duplication as well as a meaningless tag id.
That's why I gave up and decided I would rather keep the tag opaque rather than parse it incorrectly.
This PR is attempting to fix #80 by adding a codec for the
PrintImageMatchingtag. I tried to find more information about this tag structure, but it was not an easy task. Here is what I was able to find:I hope this PR makes sense. It definitely unblocks my use case - I am trying to read and write basic exif tags like
DateTimeOriginalandDateTimeDigitizedand was getting theErrUnparseableValuementioned in the issue. With this codec the error is gone and exif tool proves that thePrintImageMatchingtag is not tampered with or removed.