fix_929_exif2.32 (0.27->master) #1389

clanmills · 2020-11-07T20:38:36Z

See #929. Adding Exif 2.31 (2016) and Exif 2.32 support (2019).

kmilos · 2020-11-09T14:17:47Z

Ah, I just realized the easyacces API in #1385 should've also mentioned the time zone offsets, sorry I missed that.

Exif's OffsetTimeOriginal and OffsetTime are somehow equivalent to TIFF/EP TimeZoneOffset, but differently encoded: ASCII (HH:MM) for the former, but an array of ints for the latter (original being first; note that it doesn't support a couple of existing non-integer offset time zones):

Type	Index	Description
SSHORT	0	Time Zone Offset (in hours) of DateTimeOriginal tag-value relative to Greenwich Mean Time
SSHORT	1	If present, Time Zone Offset (in hours) of DateTime tag-value relative to Greenwich Mean Time

I have no idea ATM on how to handle this case in the easyaccess API, as an iterator to the entire Metadatum is expected to be returned, not a specific component of it...

clanmills · 2020-11-09T14:42:09Z

Relax. I have two immediate thoughts about this:

I think you are saying the that easyaccess doesn't deal with TimeZones. Does it need to?
I invented some machinery in 2019 to "manufacture" one or more metadatum in readMetadata() under some circumstances. I've forgotten the details, however that may be what's we need here. So when TimeZoneOffset is in the file, we manufacture OffsetTime. writeMedata() knows to convert it back and writes TimeZoneOffset into the file. With that magic, you don't need an easyaccess method because you would always look for OffsetTime!

clanmills · 2020-11-09T14:59:40Z

It could be even easier. According to the Wiki, the following the easyaccess function are: https://github.com/Exiv2/exiv2/wiki/EasyAccess-API

a-e	e-f	i-m	m-s	s-w
afPoint apertureValue brightnessValue contrast dateTimeOriginal exposureBiasValue exposureIndex	exposureMode exposureTime flash flashBias flashEnergy fNumber focalLength	imageQuality isoSpeed lensName lightSource macroMode make maxApertureValue	meteringMode model orientation saturation sceneCaptureType sceneMode sensingMethod	serialNumber sharpness shutterSpeedValue subjectArea subjectDistance whiteBalance

If you want add timeZoneOffset (or something) it can look look for OffsetTime, then TimeZoneOffset. The caller will have to inspect the type/count of the returned pointer, as it could be ascii or shorts.

clanmills · 2020-11-09T15:58:16Z

I've had another thought. EasyAccess is intended to make it easy to access some metadata. It's not easy to access the TZ information! So, the easyaccess API in #1385 works as documented in the Wiki.

With this PR, Exiv2 provides support for Exif 2.31 tags OffsetTime +Original +Digitized.

Exiv2 has supported the TIFF/EF tag TimeZoneOffset for sometime.

kmilos · 2020-11-09T16:06:40Z

EasyAccess is intended to make it easy to access some metadata. It's not easy to access the TZ information!

Indeed, so I think it's best left out for now.

clanmills · 2020-11-11T13:05:19Z

I had a totally new thought about this while running in the forest this morning (good place to think).

We could have another 'layer' of code in Exiv2 between the library and applications in the file src/exiv2glue.cpp. We could have define APIs such as Exiv2Glue::getTZInfo() and the like. We add unit tests for those things.

In this way we could add/modify APIs without changing the library API. This is C++ code. If users want to use Exiv2Glue functions, they have to compile/link src/exiv2glue.cpp with their application.

kmilos · 2020-11-11T18:34:31Z

Nice idea! Though since it's a departure from the "established" 0.27 features, I'd write it down on the 0.28 wishlist? In any case, nobody seems to be missing this ATM...

clanmills · 2020-11-15T06:43:54Z

I'd like to merge and close this. Please accept the invitation from GitHub to join Team Exiv2 and I can assign you as the reviewer. Last week, I asked Luis to review this, however nothing has happened. I can disable the setting to "require approvals", however it's better that changes are approved.

kmilos · 2020-11-15T14:44:07Z

src/tags_int.cpp

@@ -834,6 +834,12 @@ namespace Exiv2 {
                "version of a TIFF/EP file, eg '1', '0', '0', '0'"),
                ifd0Id, tiffEp, unsignedByte, 4, printValue), // TIFF/EP Tag
        TagInfo(0x9217, "SensingMethod", N_("Sensing Method"), N_("Type of image sensor."), ifd0Id, tiffEp, unsignedShort, 1, printValue), // TIFF/EP tag
+        TagInfo(0x9400, "Temperature", N_("Temperature"), N_("Shooting situation."), ifd0Id, otherTags, signedRational, 1, printValue), // Exif 2.31


Tags 0x9400 through to 0x9405 should be moved to Exif only section, and be part of exifId (not ifd0Id), like the others.

Also some comments should be fixed: 2 "TExif" typos, and 0x9405 is also "Exif 2.31" and not "TIFF/EP"

I don't think the 0x9217 tag is part of this PR. I agree that the comment is wrong.

Yes. The Exif2.3 tags should be in exifId.

I've been distracted by the tiffEp,otherTags and other values. I can't remember the purpose of that field. The comment in tags_int.hpp says:

/*! @brief Section identifiers to logically group tags. A section consists of nothing more than a name, based on the Exif standard. */ enum SectionId { sectionIdNotSet, imgStruct, recOffset, imgCharacter, otherTags, exifFormat, exifVersion, imgConfig, userInfo, relatedFile, dateTime, captureCond, gpsTags, iopTags, mpfTags, makerTags, dngTags, panaRaw, tiffEp, tiffPm6, adobeOpi, lastSectionId };

kmilos · 2020-11-15T14:46:04Z

src/tags_int.cpp

@@ -1925,6 +1949,9 @@ namespace Exiv2 {
        TagInfo(0x001e, "GPSDifferential", N_("GPS Differential"),
                N_("Indicates whether differential correction is applied to the GPS receiver."),
                gpsId, gpsTags, unsignedShort, 1, print0x001e),
+        TagInfo(0x001f, "GPSHPositioningError", N_("GPS Horizontal positioning error"),
+                N_("This tag indicates horizontal positioning errors in meters."),
+                gpsId, gpsTags, unsignedRational, 1, print0x001e),


Probably want printValue here?

kmilos · 2020-11-15T14:50:24Z

src/tags_int.cpp

+                N_("This tag indicates whether the recorded image is a composite image* or not."),
+                exifId, captureCond, unsignedShort, 1, printCompositeImage),
+        TagInfo(0xa461, "SourceImageNumberOfCompositeImage", N_("Source Image Number Of Composite Image"), // Exif 2.32
+                N_("This tag indicates the distance to the subject."), 


Description not updated?

You are right.

kmilos · 2020-11-15T14:50:28Z

src/tags_int.cpp

@@ -1752,6 +1767,15 @@ namespace Exiv2 {
                N_("This tag records the serial number of the interchangeable lens "
                "that was used in photography as an ASCII string."),
                exifId, otherTags, asciiString, 0, printValue),
+        TagInfo(0xa460, "CompositeImage", N_("Composite Image"), // Exif 2.32


Prefer comments after TagInfo() closing parenthesis for the next 3 tags, like in the rest of the code?

I'm not sure I've understood your comment.

kmilos · 2020-11-15T14:53:46Z

src/tags_int.cpp

+                N_("This tag indicates the distance to the subject."), 
+                exifId, captureCond, unsignedShort, 2, printValue),
+        TagInfo(0xa462, "SourceExposureTimesOfCompositeImage", N_("Source Exposure Times Of Composite Image"), // Exif 2.32
+                N_("This tag indicates the distance to the subject."),


Description not updated?

You are right.

kmilos · 2020-11-15T14:55:31Z

src/tags_int.cpp

+                exifId, captureCond, unsignedShort, 2, printValue),
+        TagInfo(0xa462, "SourceExposureTimesOfCompositeImage", N_("Source Exposure Times Of Composite Image"), // Exif 2.32
+                N_("This tag indicates the distance to the subject."),
+                exifId, captureCond, undefined, 1, print0xa40c),


Count should be -1?

Also, the pretty-print function is wrong (copy/paste error?), probably want printValue?

What should be -1?

src/tags_int.hpp

clanmills · 2020-11-15T15:44:40Z

@kmilos Thank you for looking at this and your insightful comments. Winding my way through your comments one at a time is difficult. Can you edit the PR, please? Or submit a new PR.

clanmills · 2020-11-15T18:59:28Z

I see what SectionId is about. In the Exif2.32 spec there's a table in 4.6.4. These are tags in IFD0.

A. Tags relating to image data structure
B. Tags relating to recording offset Image data location
C. Tags relating to image data characteristics
D. Other tags

4.6.5 has tags in the Exif IFD.

A. Tags Relating to Version Exif version
B. Tag Relating to Image Data Characteristics
C. Tags Relating to Image Configuration
D. Tags Relating to User Information Manufacturer notes
E. Tag Relating to Related File Information
F. Tags Relating to Date and Time
G. Tags Relating to Picture-Taking Conditions See Table 8
G2. Tags Relating to shooting situation
H. Other Tags

4.6.6 has tags in the GPS IFD.

4.6.7 has tags in the Interoperability IFD
I've never heard of this before.

I'll need to study this and resubmit the PR. For sure, SectionId is used sparingly in Exiv2. Perhaps it was added as "might be useful one day" and is now confusing.

592 rmills@rmillsmbp:~/gnu/github/exiv2/0.27-maintenance $ find . -name "*pp" | xargs grep SectionId
./src/tags.cpp:        { lastSectionId,   "(LastSection)",        N_("Last section")                 }
./src/tags_int.hpp:    enum SectionId { sectionIdNotSet,
./src/tags_int.hpp:                     lastSectionId };
./src/tags_int.hpp:        SectionId sectionId_;                   //!< Section id
593 rmills@rmillsmbp:~/gnu/github/exiv2/0.27-maintenance $

It must remain in the code because it's exposed in the API:

595 rmills@rmillsmbp:~/gnu/github/exiv2/0.27-maintenance $ find . -name "*pp" | xargs grep sectionName
./include/exiv2/tags.hpp:        static const char* sectionName(const ExifKey& key);
./samples/key-test.cpp:    // sectionName
./samples/key-test.cpp:    if (strcmp(ExifTags::sectionName(ek), "Interoperability") != 0) {
./samples/key-test.cpp:        std::cout << "Testcase failed (sectionName)" << std::endl;
./samples/key-test.cpp:    // sectionName
./samples/key-test.cpp:    if (strcmp(ExifTags::sectionName(ek2), "Interoperability") != 0) {
./samples/key-test.cpp:        std::cout << "Testcase failed (sectionName)" << std::endl;
./src/tags.cpp:    const char* ExifTags::sectionName(const ExifKey& key)

kmilos · 2020-11-15T19:29:22Z

I see what SectionId is about. In the Exif2.32 spec there's a table in 4.6.4. These are tags in IFD0.

That's right, but not only in IFD0, the Exif IFD is also divided in sections (Table 4.6.5). I think most of them are already correct, apart from the 5 new 2.31 tags relating to "G2 shooting situation". I'm ok with those remaining in the "otherTags" section, otherwise I think this is the only new section needed...

I assumed it was used in the code to further group tags into said sections when printing.

I'll let you have another go at the PR changes if that was your plan, and then I'll review again and make further changes directly if necessary...

clanmills · 2020-11-16T03:17:47Z

I'll fix the PR on Monday and you can look at it. I'll also make changes to the book concerning this. At the moment, the book says the subject of the Tiff tags (in the Tiff 6.0 Spec) and DNG and Exif tags requires more investigation. I'll get the PR fixed and you can review.

I've put everything I know into the book and yet new something new appears every week. For sure, the book is needed for future maintenance or to write a new metadata engine.

@brief

…the "section" enum is of much importance. I don't believe anything in particular is done with with it. ```cpp /*! @brief Section identifiers to logically group tags. A section consists of nothing more than a name, based on the Exif standard. */ enum SectionId { sectionIdNotSet, imgStruct, recOffset, imgCharacter, otherTags, exifFormat, exifVersion, imgConfig, userInfo, relatedFile, dateTime, captureCond, gpsTags, iopTags, mpfTags, makerTags, dngTags, panaRaw, tiffEp, tiffPm6, adobeOpi, lastSectionId }; ``` I pleased with the documentation and drawing I have added to the book concerning the Exif Specification. https://clanmills.com/exiv2/book/#Exif

…/approve.

kmilos · 2020-11-20T12:39:43Z

src/tags_int.cpp

@@ -833,7 +833,13 @@ namespace Exiv2 {
                N_("Contains four ASCII characters representing the TIFF/EP standard "
                "version of a TIFF/EP file, eg '1', '0', '0', '0'"),
                ifd0Id, tiffEp, unsignedByte, 4, printValue), // TIFF/EP Tag
-        TagInfo(0x9217, "SensingMethod", N_("Sensing Method"), N_("Type of image sensor."), ifd0Id, tiffEp, unsignedShort, 1, printValue), // TIFF/EP tag
+        TagInfo(0x9217, "SensingMethod", N_("Sensing Method"), N_("Type of image sensor."), exifId, tiffEp, unsignedShort, 1, printValue), // TIFF/EP tag
+        TagInfo(0x9400, "Temperature", N_("Temperature"), N_("Temperature as the ambient situation at the shot, for example the room temperature where the photographer was holding the camera. The unit is degrees C"), exifId, captureCond, signedRational, 1, printValue), // Exif 2.31


Still think tags 0x9400 to 0x9405 should be moved to exifTagInfo[] (just like OffsetTime etc.), they really are Exif specific.

Ah, yes. You are totally correct. Yes. Let's do that.

kmilos · 2020-11-20T12:42:06Z

src/tags_int.cpp

@@ -833,7 +833,13 @@ namespace Exiv2 {
                N_("Contains four ASCII characters representing the TIFF/EP standard "
                "version of a TIFF/EP file, eg '1', '0', '0', '0'"),
                ifd0Id, tiffEp, unsignedByte, 4, printValue), // TIFF/EP Tag
-        TagInfo(0x9217, "SensingMethod", N_("Sensing Method"), N_("Type of image sensor."), ifd0Id, tiffEp, unsignedShort, 1, printValue), // TIFF/EP tag
+        TagInfo(0x9217, "SensingMethod", N_("Sensing Method"), N_("Type of image sensor."), exifId, tiffEp, unsignedShort, 1, printValue), // TIFF/EP tag


0x9217 should stay associated w/ ifd0Id, this one is TIFF/EP specific. There is a separate (and equivalent) 0xa217 for the Exif IFD.

~~But this one should probably also use print0xa217.~~ Slightly different unfortunately, we need new translation values :(

OK, let's use print0xa217().

kmilos · 2020-11-20T12:46:31Z

src/tags_int.cpp

+        { 3, N_("CompositeCapturedWhenShooting")   }
+    };
+
+    std::ostream& printCompositeImage(std::ostream& os, const Value& value, const ExifData* metadata)


Let's use the macro EXV_PRINT_TAG() instead, no need to clutter the API with functions that are almost never used...

I agree. The code was cut'n'pasted without really thinking/understanding how everything fits together. We shouldn't add function to the API without good reason.

clanmills · 2020-11-20T12:59:50Z

@kmilos Thanks very much for the thoughtful review. 2 heads are much better than one. In software (as in life) 1+1 == 3. Can you edit the PR and let's see the back of Exif 2.32 support.

It has been a very successful and productive week. Best week I've had (both on the book and the project) since mid-June when I got tvisitor.cpp working. Very optimistic that the book will finished by the end of the year.

kmilos · 2020-11-20T13:34:34Z

Ok, I'll push my proposal in the coming days when I find some time.

Btw, do you maybe have access to the official, published TIFF/EP spec to confirm the difference between TIFF/EP 0x9217 and Exif 0xa217 values? I only have the draft... ExifTool also shows a difference btw, so I'll assume it's real...

clanmills · 2020-11-20T13:58:17Z

TIFF-EP

I have ISO/DIS 12234-2 which doesn't look like a draft. I got it from here: http://www.barrypearson.co.uk/top2009/downloads.htm

5.2.16 SensingMethod
This mandatory tag encodes the type of image sensor used in the camera or image capturing device.
Tag Name = SensingMethod Tag = 37399 (9217.H)
Type = SHORT
N=1
Value = VALUE (see below)

0 Undefined
1 MonochromeArea sensor
2 OneChipColorArea sensor
3 TwoChipColorArea sensor
4 ThreeChipColorArea sensor
5 ColorSequentialArea sensor
6 MonochromeLinear sensor
7 TriLinear sensor
8 ColorSequentialLinear sensor

Exif 2.32

I have CIPA DC-008-2019 which also doesn't appear to be a draft. I got it here. https://fotomagazin.hu/wp-content/uploads/2020/05/CIPA_DC_008_EXIF_2019.pdf

SensingMethod
Indicates the image sensor type on the camera or input device. The values are as follows.
Tag = 41495 (A217.H)
Type = SHORT
Count = 1
Default = None
1 = None
2 = One-chip color area sensor
3 = Two-chip color area sensor
4 = Three-chip color area sensor
5 = Color sequential area sensor
7 = Trilinear sensor
8 = Color sequential linear sensor reserved
Other = reserved

Difference

Caution I am dyslexic.

TIFF-EF defines value 6.

print0x217() is in the API.

952 rmills@rmillsmm-local:~/gnu/github/exiv2/0.27-maintenance $ nm -g --demangle build/lib/libexiv2.dylib  | grep 217 | cut '-d ' -f 3- | grep 217
Exiv2::Internal::print0xa217(std::__1::basic_ostream<char, std::__1::char_traits<char> >&, Exiv2::Value const&, Exiv2::ExifData const*)
953 rmills@rmillsmm-local:~/gnu/github/exiv2/0.27-maintenance $

kmilos · 2020-11-20T14:08:49Z

Thanks for checking. That is the same draft I have (don't think Mr Pearson would be allowed to distribute a published ISO standard), so we'll go with that. It's not only addition of 6 in TIFF/EP, but all values below 6 are shifted down by one as well...

clanmills · 2020-11-20T14:16:49Z

Yes. They are similar and different. Don't change the API (keep print0xa217) and use the macro EXV_PRINT_TAG() for 0x9217. They are different tags. Similar, however different.

The nice thing about standards is there's plenty of choice! You also have to pay for them because it's the only source of revenue for most standards organisations.

Have a nice weekend and thank you for your contributions.

Handle more TIFF/EP vs Exif differences Moved translated values (GPS, Exif) closer to their respective lists

kmilos · 2020-11-20T20:26:06Z

Let me know what you think about this last commit. I also took to opportunity to iron out some other minor TIFF/EP vs Exif differences.

clanmills · 2020-11-21T10:19:35Z

@kmilos Thank You Miloš for your work on this. For sure that code's in better shape today than it was last weekend. Thank You for working on this. I moved an orphaned declaration to be with his buddies.

kmilos · 2020-11-21T12:15:21Z

Thanks, looks good. Think we can merge and be done with Exif 2.32 support, otherwise we can keep prettifying the codebase till the cows come home...

clanmills · 2020-11-21T12:48:24Z

We move on, or we'll sink in the mud!

fix_929_exif2.31_0.27

58ae0b0

clanmills added this to the v0.27.4 milestone Nov 7, 2020

clanmills requested a review from piponazo November 7, 2020 20:38

clanmills self-assigned this Nov 7, 2020

clanmills linked an issue Nov 7, 2020 that may be closed by this pull request

Add support for Exif 2.32 #929

Closed

kmilos requested changes Nov 15, 2020

View reviewed changes

clanmills mentioned this pull request Nov 17, 2020

Exif 2.32 Time Zone field names are missing #1396

Closed

clanmills linked an issue Nov 18, 2020 that may be closed by this pull request

Exif 2.32 Time Zone field names are missing #1396

Closed

clanmills added 3 commits November 19, 2020 15:40

Fix tag GPSHPositioningError to use printValue() pretty-printer.

4c52891

Clarified definition and use of enum SectionId @kmilos: please review…

829ed7d

…/approve.

kmilos requested changes Nov 20, 2020

View reviewed changes

Moved ambient tags to Exif only list, other refactoring

c688ed4

Handle more TIFF/EP vs Exif differences Moved translated values (GPS, Exif) closer to their respective lists

Moved orphaned declaration of exifGPSDirRef.

5f9daf7

kmilos approved these changes Nov 21, 2020

View reviewed changes

clanmills merged commit e045fd6 into 0.27-maintenance Nov 21, 2020

clanmills deleted the fix_929_exif2.31_0.27 branch November 21, 2020 12:47

clanmills changed the title ~~fix_929_exif2.32_0.27~~ fix_929_exif2.32 (0.27->master) Nov 26, 2020

clanmills mentioned this pull request Jan 11, 2021

Exiv2 RoadMap #1018

Open

fix_929_exif2.32 (0.27->master) #1389

fix_929_exif2.32 (0.27->master) #1389

Conversation

clanmills commented Nov 7, 2020

kmilos commented Nov 9, 2020

clanmills commented Nov 9, 2020

clanmills commented Nov 9, 2020

clanmills commented Nov 9, 2020

kmilos commented Nov 9, 2020

clanmills commented Nov 11, 2020

kmilos commented Nov 11, 2020 • edited Loading

clanmills commented Nov 15, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

clanmills commented Nov 15, 2020

clanmills commented Nov 15, 2020

kmilos commented Nov 15, 2020 • edited Loading

clanmills commented Nov 16, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kmilos Nov 20, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

clanmills commented Nov 20, 2020

kmilos commented Nov 20, 2020

clanmills commented Nov 20, 2020

kmilos commented Nov 20, 2020

clanmills commented Nov 20, 2020

kmilos commented Nov 20, 2020

clanmills commented Nov 21, 2020

kmilos commented Nov 21, 2020

clanmills commented Nov 21, 2020

kmilos commented Nov 11, 2020 •

edited

Loading

kmilos commented Nov 15, 2020 •

edited

Loading

kmilos Nov 20, 2020 •

edited

Loading