123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434 |
- /* ====================================================================
- Licensed to the Apache Software Foundation (ASF) under one or more
- contributor license agreements. See the NOTICE file distributed with
- this work for additional information regarding copyright ownership.
- The ASF licenses this file to You under the Apache License, Version 2.0
- (the "License"); you may not use this file except in compliance with
- the License. You may obtain a copy of the License at
-
- http://www.apache.org/licenses/LICENSE-2.0
-
- Unless required by applicable law or agreed to in writing, software
- distributed under the License is distributed on an "AS IS" BASIS,
- WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- See the License for the specific language governing permissions and
- limitations under the License.
- ==================================================================== */
- package org.apache.poi;
-
-
- import static org.junit.Assert.assertFalse;
- import static org.junit.Assert.assertNotNull;
-
- import java.io.BufferedInputStream;
- import java.io.File;
- import java.io.FileInputStream;
- import java.io.InputStream;
- import java.util.ArrayList;
- import java.util.Arrays;
- import java.util.Collections;
- import java.util.HashMap;
- import java.util.HashSet;
- import java.util.List;
- import java.util.Locale;
- import java.util.Map;
- import java.util.Set;
-
- import org.apache.poi.stress.AbstractFileHandler;
- import org.apache.poi.stress.FileHandler;
- import org.apache.poi.stress.HDGFFileHandler;
- import org.apache.poi.stress.HMEFFileHandler;
- import org.apache.poi.stress.HPBFFileHandler;
- import org.apache.poi.stress.HPSFFileHandler;
- import org.apache.poi.stress.HSLFFileHandler;
- import org.apache.poi.stress.HSMFFileHandler;
- import org.apache.poi.stress.HSSFFileHandler;
- import org.apache.poi.stress.HWPFFileHandler;
- import org.apache.poi.stress.OPCFileHandler;
- import org.apache.poi.stress.POIFSFileHandler;
- import org.apache.poi.stress.XDGFFileHandler;
- import org.apache.poi.stress.XSLFFileHandler;
- import org.apache.poi.stress.XSSFBFileHandler;
- import org.apache.poi.stress.XSSFFileHandler;
- import org.apache.poi.stress.XWPFFileHandler;
- import org.apache.tools.ant.DirectoryScanner;
- import org.junit.Test;
- import org.junit.runner.RunWith;
- import org.junit.runners.Parameterized;
- import org.junit.runners.Parameterized.Parameter;
- import org.junit.runners.Parameterized.Parameters;
-
- /**
- * This is an integration test which performs various actions on all stored test-files and tries
- * to reveal problems which are introduced, but not covered (yet) by unit tests.
- *
- * This test looks for any file under the test-data directory and tries to do some useful
- * processing with it based on it's type.
- *
- * The test is implemented as a junit {@link Parameterized} test, which leads
- * to one test-method call for each file (currently around 950 files are handled).
- *
- * There is a a mapping of extension to implementations of the interface
- * {@link FileHandler} which defines how the file is loaded and which actions are
- * tried with the file.
- *
- * The test can be expanded by adding more actions to the FileHandlers, this automatically
- * applies the action to any such file in our test-data repository.
- *
- * There is also a list of files that should actually fail.
- *
- * Note: It is also a test-failure if a file that is expected to fail now actually works,
- * i.e. if a bug was fixed in POI itself, the file should be removed from the expected-failures
- * here as well! This is to ensure that files that should not work really do not work, e.g.
- * that we do not remove expected sanity checks.
- */
- @RunWith(Parameterized.class)
- public class TestAllFiles {
-
- private static final File ROOT_DIR = new File("test-data");
-
- static final String[] SCAN_EXCLUDES = new String[] { "**/.svn/**", "lost+found" };
-
- // map file extensions to the actual mappers
- static final Map<String, FileHandler> HANDLERS = new HashMap<String, FileHandler>();
- static {
- // Excel
- HANDLERS.put(".xls", new HSSFFileHandler());
- HANDLERS.put(".xlsx", new XSSFFileHandler());
- HANDLERS.put(".xlsm", new XSSFFileHandler());
- HANDLERS.put(".xltx", new XSSFFileHandler());
- HANDLERS.put(".xlsb", new XSSFBFileHandler());
-
- // Word
- HANDLERS.put(".doc", new HWPFFileHandler());
- HANDLERS.put(".docx", new XWPFFileHandler());
- HANDLERS.put(".dotx", new XWPFFileHandler());
- HANDLERS.put(".docm", new XWPFFileHandler());
-
- // OpenXML4J files
- HANDLERS.put(".ooxml", new OPCFileHandler()); // OPCPackage
- HANDLERS.put(".zip", new OPCFileHandler()); // OPCPackage
-
- // Powerpoint
- HANDLERS.put(".ppt", new HSLFFileHandler());
- HANDLERS.put(".pptx", new XSLFFileHandler());
- HANDLERS.put(".pptm", new XSLFFileHandler());
- HANDLERS.put(".ppsm", new XSLFFileHandler());
- HANDLERS.put(".ppsx", new XSLFFileHandler());
- HANDLERS.put(".thmx", new XSLFFileHandler());
- HANDLERS.put(".potx", new XSLFFileHandler());
-
- // Outlook
- HANDLERS.put(".msg", new HSMFFileHandler());
-
- // Publisher
- HANDLERS.put(".pub", new HPBFFileHandler());
-
- // Visio - binary
- HANDLERS.put(".vsd", new HDGFFileHandler());
-
- // Visio - ooxml
- HANDLERS.put(".vsdm", new XDGFFileHandler());
- HANDLERS.put(".vsdx", new XDGFFileHandler());
- HANDLERS.put(".vssm", new XDGFFileHandler());
- HANDLERS.put(".vssx", new XDGFFileHandler());
- HANDLERS.put(".vstm", new XDGFFileHandler());
- HANDLERS.put(".vstx", new XDGFFileHandler());
-
- // Visio - not handled yet
- HANDLERS.put(".vst", new NullFileHandler());
- HANDLERS.put(".vss", new NullFileHandler());
-
- // POIFS
- HANDLERS.put(".ole2", new POIFSFileHandler());
-
- // Microsoft Admin Template?
- HANDLERS.put(".adm", new HPSFFileHandler());
-
- // Microsoft TNEF
- HANDLERS.put(".dat", new HMEFFileHandler());
-
- // TODO: are these readable by some of the formats?
- HANDLERS.put(".shw", new NullFileHandler());
- HANDLERS.put(".zvi", new NullFileHandler());
- HANDLERS.put(".mpp", new NullFileHandler());
- HANDLERS.put(".qwp", new NullFileHandler());
- HANDLERS.put(".wps", new NullFileHandler());
- HANDLERS.put(".bin", new NullFileHandler());
- HANDLERS.put(".xps", new NullFileHandler());
- HANDLERS.put(".sldprt", new NullFileHandler());
- HANDLERS.put(".mdb", new NullFileHandler());
- HANDLERS.put(".vml", new NullFileHandler());
-
- // ignore some file types, images, other formats, ...
- HANDLERS.put(".txt", new NullFileHandler());
- HANDLERS.put(".pdf", new NullFileHandler());
- HANDLERS.put(".rtf", new NullFileHandler());
- HANDLERS.put(".gif", new NullFileHandler());
- HANDLERS.put(".html", new NullFileHandler());
- HANDLERS.put(".png", new NullFileHandler());
- HANDLERS.put(".wmf", new NullFileHandler());
- HANDLERS.put(".emf", new NullFileHandler());
- HANDLERS.put(".dib", new NullFileHandler());
- HANDLERS.put(".svg", new NullFileHandler());
- HANDLERS.put(".pict", new NullFileHandler());
- HANDLERS.put(".jpg", new NullFileHandler());
- HANDLERS.put(".jpeg", new NullFileHandler());
- HANDLERS.put(".tif", new NullFileHandler());
- HANDLERS.put(".tiff", new NullFileHandler());
- HANDLERS.put(".wav", new NullFileHandler());
- HANDLERS.put(".pfx", new NullFileHandler());
- HANDLERS.put(".xml", new NullFileHandler());
- HANDLERS.put(".csv", new NullFileHandler());
- HANDLERS.put(".ods", new NullFileHandler());
- // VBA source files
- HANDLERS.put(".vba", new NullFileHandler());
- HANDLERS.put(".bas", new NullFileHandler());
- HANDLERS.put(".frm", new NullFileHandler());
- HANDLERS.put(".frx", new NullFileHandler()); //binary
- HANDLERS.put(".cls", new NullFileHandler());
-
- // map some files without extension
- HANDLERS.put("spreadsheet/BigSSTRecord", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecord2", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecord2CR1", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecord2CR2", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecord2CR3", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecord2CR4", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecord2CR5", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecord2CR6", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecord2CR7", new NullFileHandler());
- HANDLERS.put("spreadsheet/BigSSTRecordCR", new NullFileHandler());
- HANDLERS.put("spreadsheet/test_properties1", new NullFileHandler());
- }
-
- private static Set<String> unmodifiableHashSet(String... a) {
- return Collections.unmodifiableSet(hashSet(a));
- }
- private static Set<String> hashSet(String... a) {
- return new HashSet<String>(Arrays.asList(a));
- }
-
- // Old Word Documents where we can at least extract some text
- private static final Set<String> OLD_FILES = unmodifiableHashSet(
- "document/Bug49933.doc",
- "document/Bug51944.doc",
- "document/Word6.doc",
- "document/Word6_sections.doc",
- "document/Word6_sections2.doc",
- "document/Word95.doc",
- "document/word95err.doc",
- "document/Bug60936.doc",
- "document/Bug60942.doc",
- "document/Bug60942b.doc",
- "hpsf/TestMickey.doc",
- "document/52117.doc"
- );
-
- private static final Set<String> EXPECTED_FAILURES = unmodifiableHashSet(
- // password protected files
- "spreadsheet/password.xls",
- "spreadsheet/protected_passtika.xlsx",
- "spreadsheet/51832.xls",
- "document/PasswordProtected.doc",
- "slideshow/Password_Protected-hello.ppt",
- "slideshow/Password_Protected-56-hello.ppt",
- "slideshow/Password_Protected-np-hello.ppt",
- "slideshow/cryptoapi-proc2356.ppt",
- //"document/bug53475-password-is-pass.docx",
- //"document/bug53475-password-is-solrcell.docx",
- "spreadsheet/xor-encryption-abc.xls",
- "spreadsheet/35897-type4.xls",
- //"poifs/protect.xlsx",
- //"poifs/protected_sha512.xlsx",
- //"poifs/extenxls_pwd123.xlsx",
- //"poifs/protected_agile.docx",
- "spreadsheet/58616.xlsx",
- "poifs/60320-protected.xlsx",
-
- // TODO: fails XMLExportTest, is this ok?
- "spreadsheet/CustomXMLMapping-singleattributenamespace.xlsx",
- "spreadsheet/55864.xlsx",
- "spreadsheet/57890.xlsx",
-
- // TODO: these fail now with some NPE/file read error because we now try to compute every value via Cell.toString()!
- "spreadsheet/44958.xls",
- "spreadsheet/44958_1.xls",
- "spreadsheet/testArraysAndTables.xls",
-
- // TODO: good to ignore?
- "spreadsheet/sample-beta.xlsx",
- "document/cpansearch.perl.org_src_tobyink_acme-rundoc-0.001_word-lib_hello_world.docm",
-
- // This is actually a spreadsheet!
- "hpsf/TestRobert_Flaherty.doc",
-
- // some files that are broken, eg Word 95, ...
- "spreadsheet/43493.xls",
- "spreadsheet/46904.xls",
- "document/Bug50955.doc",
- "document/57843.doc",
- "slideshow/PPT95.ppt",
- "openxml4j/OPCCompliance_CoreProperties_DCTermsNamespaceLimitedUseFAIL.docx",
- "openxml4j/OPCCompliance_CoreProperties_DoNotUseCompatibilityMarkupFAIL.docx",
- "openxml4j/OPCCompliance_CoreProperties_LimitedXSITypeAttribute_NotPresentFAIL.docx",
- "openxml4j/OPCCompliance_CoreProperties_LimitedXSITypeAttribute_PresentWithUnauthorizedValueFAIL.docx",
- "openxml4j/OPCCompliance_CoreProperties_OnlyOneCorePropertiesPartFAIL.docx",
- "openxml4j/OPCCompliance_CoreProperties_UnauthorizedXMLLangAttributeFAIL.docx",
- "openxml4j/OPCCompliance_DerivedPartNameFAIL.docx",
- "openxml4j/invalid.xlsx",
- "spreadsheet/54764-2.xlsx", // see TestXSSFBugs.bug54764()
- "spreadsheet/54764.xlsx", // see TestXSSFBugs.bug54764()
- "poifs/unknown_properties.msg", // POIFS properties corrupted
- "poifs/only-zero-byte-streams.ole2", // No actual contents
- "spreadsheet/poc-xmlbomb.xlsx", // contains xml-entity-expansion
- "spreadsheet/poc-xmlbomb-empty.xlsx", // contains xml-entity-expansion
- "spreadsheet/poc-shared-strings.xlsx", // contains shared-string-entity-expansion
- "spreadsheet/60255_extra_drawingparts.xlsx", // Non-drawing drawing
-
- // old Excel files, which we only support simple text extraction of
- "spreadsheet/testEXCEL_2.xls",
- "spreadsheet/testEXCEL_3.xls",
- "spreadsheet/testEXCEL_4.xls",
- "spreadsheet/testEXCEL_5.xls",
- "spreadsheet/testEXCEL_95.xls",
- "spreadsheet/59074.xls",
- "spreadsheet/60284.xls",
-
- // OOXML Strict is not yet supported, see bug #57699
- "spreadsheet/SampleSS.strict.xlsx",
- "spreadsheet/SimpleStrict.xlsx",
- "spreadsheet/sample.strict.xlsx",
- "spreadsheet/57914.xlsx",
-
- // non-TNEF files
- "ddf/Container.dat",
- "ddf/47143.dat",
-
- // sheet cloning errors
- "spreadsheet/47813.xlsx",
- "spreadsheet/56450.xls",
- "spreadsheet/OddStyleRecord.xls",
- "spreadsheet/WithChartSheet.xlsx",
- "spreadsheet/chart_sheet.xlsx",
- "spreadsheet/SimpleScatterChart.xlsx"
- );
-
- private static final Set<String> IGNORED = unmodifiableHashSet(
- // need JDK8+ - https://bugs.openjdk.java.net/browse/JDK-8038081
- "slideshow/42474-2.ppt",
- // OPC handler works / XSSF handler fails
- "spreadsheet/57181.xlsm"
- );
-
- @Parameters(name="{index}: {0} using {1}")
- public static Iterable<Object[]> files() {
- DirectoryScanner scanner = new DirectoryScanner();
- scanner.setBasedir(ROOT_DIR);
- scanner.setExcludes(SCAN_EXCLUDES);
-
- scanner.scan();
-
- System.out.println("Handling " + scanner.getIncludedFiles().length + " files");
-
- List<Object[]> files = new ArrayList<Object[]>();
- for(String file : scanner.getIncludedFiles()) {
- file = file.replace('\\', '/'); // ... failures/handlers lookup doesn't work on windows otherwise
- if (IGNORED.contains(file)) {
- System.out.println("Ignoring " + file);
- continue;
- }
- FileHandler handler = HANDLERS.get(getExtension(file));
- files.add(new Object[] { file, handler });
-
- // for some file-types also run OPCFileHandler
- if(handler instanceof XSSFFileHandler ||
- handler instanceof XWPFFileHandler ||
- handler instanceof XSLFFileHandler ||
- handler instanceof XDGFFileHandler) {
- files.add(new Object[] { file, HANDLERS.get(".ooxml") });
- }
- }
-
- return files;
- }
-
- @SuppressWarnings("DefaultAnnotationParam")
- @Parameter(value=0)
- public String file;
-
- @Parameter(value=1)
- public FileHandler handler;
-
- @Test
- public void testAllFiles() throws Exception {
- System.out.println("Reading " + file + " with " + handler.getClass());
- assertNotNull("Unknown file extension for file: " + file + ": " + getExtension(file), handler);
- File inputFile = new File(ROOT_DIR, file);
-
- try {
- InputStream stream = new BufferedInputStream(new FileInputStream(inputFile), 64*1024);
- try {
- handler.handleFile(stream);
-
- assertFalse("Expected to fail for file " + file + " and handler " + handler + ", but did not fail!",
- OLD_FILES.contains(file));
- } finally {
- stream.close();
- }
-
- handler.handleExtracting(inputFile);
-
- // special cases where docx-handling breaks, but OPCPackage handling works
- boolean ignoredOPC = (file.endsWith(".docx") || file.endsWith(".xlsx") ||
- file.endsWith(".xlsb") || file.endsWith(".pptx")) &&
- handler instanceof OPCFileHandler;
-
- assertFalse("Expected to fail for file " + file + " and handler " + handler + ", but did not fail!",
- EXPECTED_FAILURES.contains(file) && !ignoredOPC);
- } catch (OldFileFormatException e) {
- // for old word files we should still support extracting text
- if(OLD_FILES.contains(file)) {
- handler.handleExtracting(inputFile);
- } else {
- // check if we expect failure for this file
- if(!EXPECTED_FAILURES.contains(file) && !AbstractFileHandler.EXPECTED_EXTRACTOR_FAILURES.contains(file)) {
- System.out.println("Failed: " + file);
- throw new Exception("While handling " + file, e);
- }
- }
- } catch (Exception e) {
- // check if we expect failure for this file
- if(!EXPECTED_FAILURES.contains(file) && !AbstractFileHandler.EXPECTED_EXTRACTOR_FAILURES.contains(file)) {
- System.out.println("Failed: " + file);
- throw new Exception("While handling " + file, e);
- }
- }
-
- // let some file handlers do additional stuff
- handler.handleAdditional(inputFile);
- }
-
- static String getExtension(String file) {
- int pos = file.lastIndexOf('.');
- if(pos == -1 || pos == file.length()-1) {
- return file;
- }
-
- return file.substring(pos).toLowerCase(Locale.ROOT);
- }
-
- private static class NullFileHandler implements FileHandler {
- @Override
- public void handleFile(InputStream stream) throws Exception {
- }
-
- @Override
- public void handleExtracting(File file) throws Exception {
- }
-
- @Override
- public void handleAdditional(File file) throws Exception {
- }
- }
- }
|