Improved handling of corrupted JPEG (#91)

Added some broken JPEG files to the testing lib and changed logic to
read some of them.

README.md

@@ -13,7 +13,7 @@
 
 Kim is a Kotlin Multiplatform library for reading and writing image metadata.
 
-It's part of [Ashampoo Photos](https://ashampoo.com/photos).
+It's part of [Ashampoo Photo Organizer](https://ashampoo.com/photo-organizer).
 
 ## Features
 
@@ -34,12 +34,12 @@ It's part of [Ashampoo Photos](https://ashampoo.com/photos).
     + JPG: Lossless rotation by modifying only one byte (where present)
 
 The future development of features on our part is driven entirely by the needs
-of Ashampoo Photos, which, in turn, is driven by user community feedback.
+of Ashampoo Photo Organizer, which, in turn, is driven by user community feedback.
 
 ## Installation
 
 ```
-implementation("com.ashampoo:kim:0.17.3")
+implementation("com.ashampoo:kim:0.17.4")
 ```
 
 For the targets `wasmJs` & `js` you also need to specify this:

examples/kim-kotlin-jvm-sample/src/main/kotlin/Main.kt

@@ -125,6 +125,8 @@ fun setGeoTiffToJpeg() {
 
 /**
  * Shows how to update set GeoTiff to a TIF file using JVM API.
+ *
+ * CAUTION: Writing TIFF is experimental and may corrupt the file!
  */
 fun setGeoTiffToTiff() {
 

src/commonMain/kotlin/com/ashampoo/kim/common/ExifDateUtil.kt

@@ -15,13 +15,9 @@
  */
 package com.ashampoo.kim.common
 
-/**
- * This class contains extensions needed to
- * interpret bytes in photo files.
- */
-
-private val emptyExifDates = setOf(
+private val emptyExifDateStrings = setOf(
     "0000:00:00 00:00:00",
+    "    :  :     :  :  ",
     "                   "
 )
 
@@ -43,7 +39,7 @@ private const val FIRST_SECOND_INDEX = 17
 private const val SECOND_SECOND_INDEX = 18
 
 fun isExifDateEmpty(exifDate: String?): Boolean =
-    exifDate.isNullOrEmpty() || emptyExifDates.contains(exifDate)
+    exifDate.isNullOrBlank() || emptyExifDateStrings.contains(exifDate)
 
 /**
  * EXIF dates are in the format of "yyyy:MM:dd HH:mm:ss" (19 chars),

src/commonMain/kotlin/com/ashampoo/kim/common/PhotoMetadataConverter.kt

@@ -31,167 +31,187 @@ import com.ashampoo.kim.model.TiffOrientation
 import kotlinx.datetime.LocalDateTime
 import kotlinx.datetime.TimeZone
 import kotlinx.datetime.toInstant
+import kotlin.jvm.JvmOverloads
+import kotlin.jvm.JvmStatic
 
-fun ImageMetadata.convertToPhotoMetadata(
-    ignoreOrientation: Boolean = false
-): PhotoMetadata {
+/*
+ * This is a dedicated object with @JvmStatic methods
+ * to provide a better API to pure Java projects.
+ */
+object PhotoMetadataConverter {
+
+    @JvmStatic
+    @JvmOverloads
+    @Suppress("LongMethod")
+    fun convertToPhotoMetadata(
+        imageMetadata: ImageMetadata,
+        ignoreOrientation: Boolean = false
+    ): PhotoMetadata {
+
+        val orientation = if (ignoreOrientation)
+            TiffOrientation.STANDARD
+        else
+            TiffOrientation.of(imageMetadata.findShortValue(TiffTag.TIFF_TAG_ORIENTATION)?.toInt())
 
-    val orientation = if (ignoreOrientation)
-        TiffOrientation.STANDARD
-    else
-        TiffOrientation.of(findShortValue(TiffTag.TIFF_TAG_ORIENTATION)?.toInt())
+        val takenDateMillis = extractTakenDateMillis(imageMetadata)
 
-    val takenDateMillis = extractTakenDateMillis(this)
+        val gpsDirectory = imageMetadata.findTiffDirectory(TiffConstants.TIFF_DIRECTORY_GPS)
 
-    val gpsDirectory = findTiffDirectory(TiffConstants.TIFF_DIRECTORY_GPS)
+        val gps = gpsDirectory?.let { GPSInfo.createFrom(it) }
 
-    val gps = gpsDirectory?.let { GPSInfo.createFrom(it) }
+        val latitude = gps?.getLatitudeAsDegreesNorth()
+        val longitude = gps?.getLongitudeAsDegreesEast()
 
-    val latitude = gps?.getLatitudeAsDegreesNorth()
-    val longitude = gps?.getLongitudeAsDegreesEast()
+        val cameraMake = imageMetadata.findStringValue(TiffTag.TIFF_TAG_MAKE)
+        val cameraModel = imageMetadata.findStringValue(TiffTag.TIFF_TAG_MODEL)
 
-    val cameraMake = findStringValue(TiffTag.TIFF_TAG_MAKE)
-    val cameraModel = findStringValue(TiffTag.TIFF_TAG_MODEL)
+        val lensMake = imageMetadata.findStringValue(ExifTag.EXIF_TAG_LENS_MAKE)
+        val lensModel = imageMetadata.findStringValue(ExifTag.EXIF_TAG_LENS_MODEL)
 
-    val lensMake = findStringValue(ExifTag.EXIF_TAG_LENS_MAKE)
-    val lensModel = findStringValue(ExifTag.EXIF_TAG_LENS_MODEL)
+        /* Look for ISO at the standard place and fall back to test RW2 logic. */
+        val iso = imageMetadata.findShortValue(ExifTag.EXIF_TAG_ISO)
+            ?: imageMetadata.findShortValue(ExifTag.EXIF_TAG_ISO_PANASONIC)
 
-    /* Look for ISO at the standard place and fall back to test RW2 logic. */
-    val iso = findShortValue(ExifTag.EXIF_TAG_ISO)
-        ?: findShortValue(ExifTag.EXIF_TAG_ISO_PANASONIC)
+        val exposureTime = imageMetadata.findDoubleValue(ExifTag.EXIF_TAG_EXPOSURE_TIME)
+        val fNumber = imageMetadata.findDoubleValue(ExifTag.EXIF_TAG_FNUMBER)
+        val focalLength = imageMetadata.findDoubleValue(ExifTag.EXIF_TAG_FOCAL_LENGTH)
 
-    val exposureTime = findDoubleValue(ExifTag.EXIF_TAG_EXPOSURE_TIME)
-    val fNumber = findDoubleValue(ExifTag.EXIF_TAG_FNUMBER)
-    val focalLength = findDoubleValue(ExifTag.EXIF_TAG_FOCAL_LENGTH)
+        val keywords = mutableSetOf<String>()
 
-    val keywords = mutableSetOf<String>()
+        val iptcRecords = imageMetadata.iptc?.records
 
-    val iptcRecords = iptc?.records
+        iptcRecords?.forEach {
 
-    iptcRecords?.forEach {
+            if (it.iptcType == IptcTypes.KEYWORDS)
+                keywords.add(it.value)
+        }
 
-        if (it.iptcType == IptcTypes.KEYWORDS)
-            keywords.add(it.value)
-    }
+        val gpsCoordinates =
+            if (latitude != null && longitude != null)
+                GpsCoordinates(
+                    latitude = latitude,
+                    longitude = longitude
+                )
+            else
+                null
 
-    val gpsCoordinates =
-        if (latitude != null && longitude != null)
-            GpsCoordinates(
-                latitude = latitude,
-                longitude = longitude
-            )
-        else
-            null
+        val xmpMetadata: PhotoMetadata? = imageMetadata.xmp?.let {
+            XmpReader.readMetadata(it)
+        }
 
-    val xmpMetadata: PhotoMetadata? = xmp?.let {
-        XmpReader.readMetadata(it)
-    }
+        val thumbnailBytes = imageMetadata.getExifThumbnailBytes()
 
-    val thumbnailBytes = getExifThumbnailBytes()
+        val thumbnailImageSize = thumbnailBytes?.let {
+            JpegImageParser.getImageSize(
+                ByteArrayByteReader(thumbnailBytes)
+            )
+        }
 
-    val thumbnailImageSize = thumbnailBytes?.let {
-        JpegImageParser.getImageSize(
-            ByteArrayByteReader(thumbnailBytes)
+        /*
+         * Embedded XMP metadata has higher priority than EXIF or IPTC
+         * for certain fields because it's the newer format. Some fields
+         * like rating, faces and persons in image are exclusive to XMP.
+         *
+         * Resolution, orientation and capture parameters (camera make,
+         * iso, exposure time, etc.) are always taken from EXIF.
+         */
+        return PhotoMetadata(
+            widthPx = imageMetadata.imageSize?.width,
+            heightPx = imageMetadata.imageSize?.height,
+            orientation = orientation,
+            takenDate = xmpMetadata?.takenDate ?: takenDateMillis,
+            gpsCoordinates = xmpMetadata?.gpsCoordinates ?: gpsCoordinates,
+            location = xmpMetadata?.location,
+            cameraMake = cameraMake,
+            cameraModel = cameraModel,
+            lensMake = lensMake,
+            lensModel = lensModel,
+            iso = iso?.toInt(),
+            exposureTime = exposureTime,
+            fNumber = fNumber,
+            focalLength = focalLength,
+            flagged = xmpMetadata?.flagged ?: false,
+            rating = xmpMetadata?.rating,
+            keywords = keywords.ifEmpty { xmpMetadata?.keywords ?: emptySet() },
+            faces = xmpMetadata?.faces ?: emptyMap(),
+            personsInImage = xmpMetadata?.personsInImage ?: emptySet(),
+            albums = xmpMetadata?.albums ?: emptySet(),
+            thumbnailImageSize = thumbnailImageSize,
+            thumbnailBytes = thumbnailBytes
         )
     }
 
-    /*
-     * Embedded XMP metadata has higher priority than EXIF or IPTC
-     * for certain fields because it's the newer format. Some fields
-     * like rating, faces and persons in image are exclusive to XMP.
-     *
-     * Resolution, orientation and capture parameters (camera make,
-     * iso, exposure time, etc.) are always taken from EXIF.
-     */
-    return PhotoMetadata(
-        widthPx = imageSize?.width,
-        heightPx = imageSize?.height,
-        orientation = orientation,
-        takenDate = xmpMetadata?.takenDate ?: takenDateMillis,
-        gpsCoordinates = xmpMetadata?.gpsCoordinates ?: gpsCoordinates,
-        location = xmpMetadata?.location,
-        cameraMake = cameraMake,
-        cameraModel = cameraModel,
-        lensMake = lensMake,
-        lensModel = lensModel,
-        iso = iso?.toInt(),
-        exposureTime = exposureTime,
-        fNumber = fNumber,
-        focalLength = focalLength,
-        flagged = xmpMetadata?.flagged ?: false,
-        rating = xmpMetadata?.rating,
-        keywords = keywords.ifEmpty { xmpMetadata?.keywords ?: emptySet() },
-        faces = xmpMetadata?.faces ?: emptyMap(),
-        personsInImage = xmpMetadata?.personsInImage ?: emptySet(),
-        albums = xmpMetadata?.albums ?: emptySet(),
-        thumbnailImageSize = thumbnailImageSize,
-        thumbnailBytes = thumbnailBytes
-    )
-}
+    @JvmStatic
+    fun extractTakenDateAsIsoString(metadata: ImageMetadata): String? {
 
-private fun extractTakenDateAsIso(metadata: ImageMetadata): String? {
+        val takenDateField = metadata.findTiffField(ExifTag.EXIF_TAG_DATE_TIME_ORIGINAL)
+            ?: return null
 
-    val takenDateField = metadata.findTiffField(ExifTag.EXIF_TAG_DATE_TIME_ORIGINAL)
+        var takenDate = takenDateField.value as? String
 
-    var takenDate = takenDateField?.value as? String
+        /*
+         * Workaround in case that it's a String array.
+         */
+        if (takenDate == null)
+            takenDate = takenDateField.toStringValue()
 
-    /*
-     * photo_53.jpg of our test data triggers a bug here.
-     * This is workaround code.
-     */
-    if (takenDate == null && takenDateField != null)
-        takenDate = takenDateField.toStringValue()
+        if (isExifDateEmpty(takenDate))
+            return null
 
-    if (takenDate == null || isExifDateEmpty(takenDate))
-        return null
+        return convertExifDateToIso8601Date(takenDate)
+    }
 
-    return convertExifDateToIso8601Date(takenDate)
-}
+    @JvmStatic
+    fun extractTakenDateMillis(metadata: ImageMetadata): Long? {
 
-private fun extractTakenDateMillis(metadata: ImageMetadata): Long? {
+        var takenDate: String? = null
 
-    val exif = metadata.exif
+        try {
 
-    if (exif == null)
-        return exif
+            takenDate = extractTakenDateAsIsoString(metadata) ?: return null
 
-    var takenDate: String? = null
+            val takenDateSubSecond = metadata
+                .findStringValue(ExifTag.EXIF_TAG_SUB_SEC_TIME_ORIGINAL)
+                ?.toIntOrNull()
+                ?: 0
 
-    try {
+            /*
+             * If the date string itself contains a sub second like "2020-08-30T18:43:00.500"
+             * this should be used. We append it, if the string does not have a dot yet.
+             */
+            val takenDatePlusSubSecond = if (!takenDate.contains('.'))
+                "$takenDate.$takenDateSubSecond"
+            else
+                takenDate
 
-        takenDate = extractTakenDateAsIso(metadata) ?: return null
+            val timeZone = if (underUnitTesting)
+                TimeZone.of("GMT+02:00")
+            else
+                TimeZone.currentSystemDefault()
 
-        val takenDateSubSecond = metadata
-            .findStringValue(ExifTag.EXIF_TAG_SUB_SEC_TIME_ORIGINAL)
-            ?.toIntOrNull()
-            ?: 0
+            return LocalDateTime.parse(takenDatePlusSubSecond)
+                .toInstant(timeZone)
+                .toEpochMilliseconds()
 
-        /*
-         * If the date string itself contains a sub second like "2020-08-30T18:43:00.500"
-         * this should be used. We append it, if the string does not have a dot yet.
-         */
-        val takenDatePlusSubSecond = if (!takenDate.contains('.'))
-            "$takenDate.$takenDateSubSecond"
-        else
-            takenDate
+        } catch (ignore: Exception) {
 
-        val timeZone = if (underUnitTesting)
-            TimeZone.of("GMT+02:00")
-        else
-            TimeZone.currentSystemDefault()
+            /*
+             * Many photos contain wrong values here. We ignore this problem and hope
+             * that another taken date source like embedded XMP has a valid date instead.
+             */
+            println("Ignore invalid EXIF DateTimeOriginal: '$takenDate'")
 
-        return LocalDateTime.parse(takenDatePlusSubSecond)
-            .toInstant(timeZone)
-            .toEpochMilliseconds()
+            return null
+        }
+    }
 
-    } catch (ignore: Exception) {
+}
 
-        /*
-         * Many photos contain wrong values here. We ignore this problem and hope
-         * that another taken date source like embedded XMP has a valid date instead.
-         */
-        println("Ignore invalid EXIF DateTimeOriginal: '$takenDate'")
+fun ImageMetadata.convertToPhotoMetadata(
+    ignoreOrientation: Boolean = false
+): PhotoMetadata =
+    PhotoMetadataConverter.convertToPhotoMetadata(
+        imageMetadata = this,
+        ignoreOrientation = ignoreOrientation
+    )
 
-        return null
-    }
-}