This is an automated email from the ASF dual-hosted git repository. tballison pushed a commit to branch TIKA-4816-metadata-key-api in repository https://gitbox.apache.org/repos/asf/tika.git
commit 0e3fe950599a828c9676d22dbcee8ace4b0cca19 Author: tallison <[email protected]> AuthorDate: Wed Aug 12 11:41:30 2026 -0400 TIKA-4816 fix EMBEDDED_EXCEPTION erasure under active metadata limiter --- .../writefilter/StandardMetadataLimiter.java | 3 +- .../writefilter/StandardMetadataLimiterTest.java | 75 ++++++++++++++++++++++ 2 files changed, 77 insertions(+), 1 deletion(-) diff --git a/tika-core/src/main/java/org/apache/tika/metadata/writefilter/StandardMetadataLimiter.java b/tika-core/src/main/java/org/apache/tika/metadata/writefilter/StandardMetadataLimiter.java index 913057a8f0..4acd9db81a 100644 --- a/tika-core/src/main/java/org/apache/tika/metadata/writefilter/StandardMetadataLimiter.java +++ b/tika-core/src/main/java/org/apache/tika/metadata/writefilter/StandardMetadataLimiter.java @@ -82,12 +82,13 @@ public class StandardMetadataLimiter implements MetadataWriteLimiter, Serializab ALWAYS_SET_FIELDS.add(AccessPermissions.EXTRACT_FOR_ACCESSIBILITY.getName()); ALWAYS_SET_FIELDS.add(Metadata.CONTENT_DISPOSITION.getName()); ALWAYS_SET_FIELDS.add(TikaCoreProperties.CONTAINER_EXCEPTION.getName()); - ALWAYS_SET_FIELDS.add(TikaCoreProperties.EMBEDDED_EXCEPTION.getName()); //Metadata.CONTENT_LOCATION? used by the html parser } static { ALWAYS_ADD_FIELDS.add(TikaCoreProperties.TIKA_PARSED_BY.getName()); + //bag-typed: one embedded-file failure per add() call must not clobber prior ones + ALWAYS_ADD_FIELDS.add(TikaCoreProperties.EMBEDDED_EXCEPTION.getName()); } private static final String METADATA_TRUNCATED_KEY = diff --git a/tika-serialization/src/test/java/org/apache/tika/metadata/writefilter/StandardMetadataLimiterTest.java b/tika-serialization/src/test/java/org/apache/tika/metadata/writefilter/StandardMetadataLimiterTest.java index 989d1cdf11..9ad85c5fb2 100644 --- a/tika-serialization/src/test/java/org/apache/tika/metadata/writefilter/StandardMetadataLimiterTest.java +++ b/tika-serialization/src/test/java/org/apache/tika/metadata/writefilter/StandardMetadataLimiterTest.java @@ -286,6 +286,81 @@ public class StandardMetadataLimiterTest extends TikaTest { } + @Test + public void testEmbeddedExceptionAddRetainsAllValues() throws Exception { + //EMBEDDED_EXCEPTION is bag-typed and in ALWAYS_ADD_FIELDS: N adds must retain + //N values, in order, even under an active limiter + Metadata metadata = filter(100, 10000, 10000, 100, + Collections.EMPTY_SET, Collections.EMPTY_SET, true); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "exception 1"); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "exception 2"); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "exception 3"); + + assertArrayEquals(new String[]{"exception 1", "exception 2", "exception 3"}, + metadata.getValues(TikaCoreProperties.EMBEDDED_EXCEPTION)); + } + + @Test + public void testEmbeddedExceptionSetStillReplaces() throws Exception { + //set() must still replace, even though add() now appends + Metadata metadata = filter(100, 10000, 10000, 100, + Collections.EMPTY_SET, Collections.EMPTY_SET, true); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "exception 1"); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "exception 2"); + metadata.set(TikaCoreProperties.EMBEDDED_EXCEPTION, "replacement"); + + assertArrayEquals(new String[]{"replacement"}, + metadata.getValues(TikaCoreProperties.EMBEDDED_EXCEPTION)); + } + + @Test + public void testEmbeddedExceptionPerValueTruncation() throws Exception { + //per-value truncation and the truncated-metadata marker must still apply, + //without erasing values already added + Metadata metadata = filter(100, 10, 100000, 100, + Collections.EMPTY_SET, Collections.EMPTY_SET, true); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "short"); + //800 utf-16 bytes, over the 300-byte always-field floor + String longValue = "x".repeat(400); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, longValue); + + String[] values = metadata.getValues(TikaCoreProperties.EMBEDDED_EXCEPTION); + assertEquals(2, values.length); + assertEquals("short", values[0]); + //truncated to the 300-byte (utf-16) always-field floor = 150 chars + assertEquals(150, values[1].length()); + assertTruncated(metadata); + } + + @Test + public void testEmbeddedExceptionSurvivesOverBudget() throws Exception { + //always-add fields must not be dropped even once the overall budget is exhausted + Metadata metadata = filter(100, 1, 1, 100, + Collections.EMPTY_SET, Collections.EMPTY_SET, true); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "e1"); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "e2"); + metadata.add(TikaCoreProperties.EMBEDDED_EXCEPTION, "e3"); + + assertArrayEquals(new String[]{"e1", "e2", "e3"}, + metadata.getValues(TikaCoreProperties.EMBEDDED_EXCEPTION)); + } + + @Test + public void testSimpleAlwaysSetFieldStillReplacesOnAdd() throws Exception { + //CONTAINER_EXCEPTION is SIMPLE (not bag) and stays in ALWAYS_SET_FIELDS: + //the limiter's add() dispatch must keep replacing, as before this fix. + //Use addTrusted() to reach the limiter directly: Metadata.add(Property,..) itself + //already throws PropertyTypeException on a 2nd add to a SIMPLE property, so that + //typed path can't otherwise exercise the limiter's repeated-add behavior. + Metadata metadata = filter(100, 10000, 10000, 100, + Collections.EMPTY_SET, Collections.EMPTY_SET, true); + metadata.addTrusted(TikaCoreProperties.CONTAINER_EXCEPTION.getName(), "container exception 1"); + metadata.addTrusted(TikaCoreProperties.CONTAINER_EXCEPTION.getName(), "container exception 2"); + + assertArrayEquals(new String[]{"container exception 2"}, + metadata.getValues(TikaCoreProperties.CONTAINER_EXCEPTION)); + } + @Test public void testExclude() throws Exception { TikaLoader loader = TikaLoader.load(getConfigPath(getClass(), "TIKA-3695-exclude.json"));
