This is an automated email from the ASF dual-hosted git repository. tballison pushed a commit to branch TIKA-4683-fix-truncated-epub in repository https://gitbox.apache.org/repos/asf/tika.git
commit e22685b3eb507cd84fb5caf587f83b788c30b46f Author: tallison <[email protected]> AuthorDate: Wed Apr 29 10:38:56 2026 -0400 improve epub handling of truncated files --- .../org/apache/tika/parser/epub/EpubParser.java | 112 ++++++++++++++++++++- .../java/org/apache/tika/parser/pkg/ZipParser.java | 15 +++ 2 files changed, 124 insertions(+), 3 deletions(-) diff --git a/tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-miscoffice-module/src/main/java/org/apache/tika/parser/epub/EpubParser.java b/tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-miscoffice-module/src/main/java/org/apache/tika/parser/epub/EpubParser.java index ae131407d0..4460946c52 100644 --- a/tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-miscoffice-module/src/main/java/org/apache/tika/parser/epub/EpubParser.java +++ b/tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-miscoffice-module/src/main/java/org/apache/tika/parser/epub/EpubParser.java @@ -37,6 +37,8 @@ import org.apache.commons.compress.archivers.zip.ZipArchiveEntry; import org.apache.commons.compress.archivers.zip.ZipFile; import org.apache.commons.io.IOUtils; import org.apache.commons.lang3.StringUtils; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; import org.xml.sax.Attributes; import org.xml.sax.ContentHandler; import org.xml.sax.SAXException; @@ -68,6 +70,8 @@ import org.apache.tika.utils.XMLReaderUtils; @TikaComponent public class EpubParser implements Parser { + private static final Logger LOG = LoggerFactory.getLogger(EpubParser.class); + /** * Serial version UID */ @@ -155,12 +159,24 @@ public class EpubParser implements Parser { throws IOException, TikaException, SAXException { String rootOPF = getRoot(zipFile, context); + LOG.trace("epub bufferedParseZipFile: rootOPF={}", rootOPF); if (rootOPF == null) { - return Collections.EMPTY_SET; + // No container.xml and no .opf — typical of truncated epubs where + // the OPF lives past the truncation point. Fall back to iterating + // the recoverable HTML/XHTML entries in stored order so we still + // emit partial content (matching 3.x's streamingParse contract), + // then throw to signal the result is incomplete. + LOG.trace("epub fallback: rootOPF=null, streaming all html entries"); + return fallbackParseAllHtmlEntries(zipFile, bodyHandler, metadata, context, + "no OPF found in (possibly truncated) container"); } ZipArchiveEntry zae = zipFile.getEntry(rootOPF); + LOG.trace("epub OPF entry: zae={} canReadEntryData={}", + zae, zae == null ? "n/a" : zipFile.canReadEntryData(zae)); if (zae == null || !zipFile.canReadEntryData(zae)) { - return Collections.EMPTY_SET; + LOG.trace("epub fallback: OPF entry missing/unreadable, streaming all html entries"); + return fallbackParseAllHtmlEntries(zipFile, bodyHandler, metadata, context, + "OPF entry missing or unreadable in (possibly truncated) container"); } try (TikaInputStream tis = TikaInputStream.get(zipFile.getInputStream(zae))) { opf.parse(tis, new DefaultHandler(), metadata, context); @@ -170,8 +186,13 @@ public class EpubParser implements Parser { try (InputStream is = zipFile.getInputStream(zae)) { XMLReaderUtils.parseSAX(is, contentOrderScraper, context); } + LOG.trace("epub OPF parsed: spine items={}, manifest entries={}", + contentOrderScraper.contentItems.size(), + contentOrderScraper.locationMap.size()); if (contentOrderScraper.contentItems.isEmpty()) { - return Collections.EMPTY_SET; + LOG.trace("epub fallback: empty spine, streaming all html entries"); + return fallbackParseAllHtmlEntries(zipFile, bodyHandler, metadata, context, + "OPF declared no spine items in (possibly truncated) container"); } String relativePath = ""; if (rootOPF.lastIndexOf("/") > -1) { @@ -182,7 +203,9 @@ public class EpubParser implements Parser { Set<String> encryptedItems = checkForDRM(zipFile); Set<String> processed = new HashSet<>(); Set<SAXException> saxExceptions = new HashSet<>(); + int spineSeen = 0, spineParsed = 0, spineMissing = 0, spineNonHtml = 0; for (String id : contentOrderScraper.contentItems) { + spineSeen++; HRefMediaPair hRefMediaPair = contentOrderScraper.locationMap.get(id); if (hRefMediaPair != null && hRefMediaPair.href != null) { //we need to test for xhtml/xml because the content parser @@ -207,18 +230,29 @@ public class EpubParser implements Parser { if (zae != null) { try (TikaInputStream tis = TikaInputStream.get(zipFile.getInputStream(zae))) { content.parse(tis, bodyHandler, metadata, context); + spineParsed++; } catch (SAXException e) { if (WriteLimitReachedException.isWriteLimitReached(e)) { throw e; } saxExceptions.add(e); + } catch (IOException ioe) { + LOG.trace("epub spine read IOException on {}: {}", path, ioe.toString()); + throw ioe; } finally { processed.add(id); } + } else { + spineMissing++; + LOG.trace("epub spine: getEntry({}) returned null (truncated?)", path); } + } else { + spineNonHtml++; } } } + LOG.trace("epub spine summary: seen={} parsed={} missing={} non-html={}", + spineSeen, spineParsed, spineMissing, spineNonHtml); //now handle embedded files EmbeddedDocumentExtractor embeddedDocumentExtractor = @@ -240,9 +274,81 @@ public class EpubParser implements Parser { for (SAXException e : saxExceptions) { throw e; } + // If spine items referenced entries not in the (possibly salvaged) + // zip — typical of truncated epubs where the OPF survived but later + // chapters didn't — throw IOException so the outer parse() flushes + // the partial content already in xhtml and signals incompleteness. + // This restores 3.x's partial-content-plus-exception contract. + if (spineMissing > 0) { + throw new IOException("EPUB: " + spineMissing + " of " + + spineSeen + " spine items missing from (possibly truncated) " + + "container; emitted " + spineParsed + " recovered chapters"); + } return encryptedItems; } + /** + * Fallback used when the OPF can't be located or parsed (typically a + * truncated epub where the OPF lives past the truncation point). + * Iterates the zip's entries in stored order and parses any HTML/XHTML/XML + * entry, mirroring 3.x's {@code streamingParse} behaviour. Throws + * IOException at the end so the outer parse() flushes the partial content + * and the caller learns that extraction was incomplete. + */ + private Set<String> fallbackParseAllHtmlEntries(ZipFile zipFile, + ContentHandler bodyHandler, + Metadata metadata, + ParseContext context, + String reason) + throws IOException, TikaException, SAXException { + // Try to recover mimetype + metadata.xml even in the fallback path, + // since they may be present even when the OPF isn't. + try { + extractMetadata(zipFile, metadata, context); + } catch (Exception e) { + LOG.trace("epub fallback: extractMetadata threw {}", e.toString()); + } + int parsed = 0; + int failed = 0; + Enumeration<ZipArchiveEntry> entries = zipFile.getEntries(); + while (entries.hasMoreElements()) { + ZipArchiveEntry entry = entries.nextElement(); + String name = entry.getName().toLowerCase(Locale.US); + if (!(name.endsWith(".xhtml") || name.endsWith(".html") + || name.endsWith(".htm") || name.endsWith(".xml"))) { + continue; + } + // Skip the OPF file if we somehow have one but it didn't parse + // upstream — body handler isn't the right place for it. + if (name.endsWith(".opf")) { + continue; + } + if (!zipFile.canReadEntryData(entry)) { + continue; + } + try (TikaInputStream tis = TikaInputStream.get(zipFile.getInputStream(entry))) { + content.parse(tis, bodyHandler, metadata, context); + parsed++; + } catch (SAXException e) { + if (WriteLimitReachedException.isWriteLimitReached(e)) { + throw e; + } + failed++; + LOG.trace("epub fallback: SAX failure on {}: {}", entry.getName(), e.toString()); + } catch (IOException e) { + failed++; + LOG.trace("epub fallback: IO failure on {}: {}", entry.getName(), e.toString()); + } + } + LOG.trace("epub fallback summary: parsed={} failed={}", parsed, failed); + // Always throw — the caller asked for an EPUB and we couldn't follow + // the spine. Partial content was emitted to xhtml; outer parse() + // flushes it. + throw new IOException("EPUB: fallback recovery (" + reason + + "); recovered " + parsed + " HTML/XHTML entries" + + (failed > 0 ? " (" + failed + " failed)" : "")); + } + private Set<String> checkForDRM(ZipFile zipFile) throws IOException, TikaException, SAXException { ZipArchiveEntry zae = zipFile.getEntry(META_INF_ENCRYPTION); diff --git a/tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-pkg-module/src/main/java/org/apache/tika/parser/pkg/ZipParser.java b/tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-pkg-module/src/main/java/org/apache/tika/parser/pkg/ZipParser.java index eec152cb7e..d206cae649 100644 --- a/tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-pkg-module/src/main/java/org/apache/tika/parser/pkg/ZipParser.java +++ b/tika-parsers/tika-parsers-standard/tika-parsers-standard-modules/tika-parser-pkg-module/src/main/java/org/apache/tika/parser/pkg/ZipParser.java @@ -475,6 +475,21 @@ public class ZipParser extends AbstractArchiveParser { return; } + // Defensive: mirror the streaming path's canReadEntryData gate so a + // truncated / unsupported entry in a salvaged ZipFile records an + // embedded-stream exception (caller-visible signal) instead of + // silently disappearing when getInputStream/parseEmbedded fail + // partway through. + if (!zipFile.canReadEntryData(entry)) { + EmbeddedDocumentUtil.recordEmbeddedStreamException( + new TikaException("Can't read archive stream (" + name + ")"), + parentMetadata); + if (name != null && !name.isEmpty()) { + xhtml.element("p", name); + } + return; + } + Metadata entryMetadata = buildEntryMetadata(entry, name, context); writeEntryXhtml(name, xhtml);
