@cloudant/couchbackup 2.9.9-SNAPSHOT.98 → 2.9.9

package/.scannerwork/scanner-report/source-67.txt

@@ -1,4 +1,4 @@
-// Copyright © 2017, 2018 IBM Corp. All rights reserved.
+// Copyright © 2017, 2021 IBM Corp. All rights reserved.
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
@@ -11,215 +11,154 @@
 // 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.
-
-/* global describe it */
 'use strict';
 
-const assert = require('assert');
-const backup = require('../app.js').backup;
-const fs = require('fs');
-const nock = require('nock');
-
-const goodUrl = 'http://localhost:5984/db';
-// The real validateArgs function of app.js isn't
-// exported - so we call the exported backup method
-// instead. We don't get as far as a real backup when
-// testing error cases. For success cases we nock the
-// goodUrl and
-const validateArgs = function(url, opts, callback) {
-  const nullStream = fs.createWriteStream('/dev/null');
-  let cb = callback;
-  if (url === goodUrl) {
-    // Nock the goodUrl
-    nock(goodUrl).head('').reply(404, { error: 'not_found', reason: 'missing' });
-    // replace the callback to handle the nock response
-    // to avoid attempting a real backup
-    cb = function(err) {
-      nullStream.end();
-      if (err.name === 'DatabaseNotFound') {
-        // This is what we expect if we reached the backup
-        // This is success for valid args cases.
-        err = null;
-      }
-      callback(err);
-    };
-  }
-  backup(url, nullStream, opts, cb);
-  return true;
-};
+const async = require('async');
+const stream = require('stream');
+const error = require('./error.js');
+const debug = require('debug')('couchbackup:writer');
 
-const stderrWriteFun = process.stderr.write;
-let capturedStderr;
+module.exports = function(db, bufferSize, parallelism, ee) {
+  const writer = new stream.Transform({ objectMode: true });
+  let buffer = [];
+  let written = 0;
+  let linenumber = 0;
 
-function captureStderr() {
-  process.stderr.write = function(string, encoding, fd) {
-    capturedStderr += string;
-  };
-}
-
-function releaseStderr() {
-  process.stderr.write = stderrWriteFun;
-  capturedStderr = null;
-}
-
-function assertErrorMessage(msg, done) {
-  return function(err, data) {
-    try {
-      assert(err.message, 'There should be an error message');
-      assert(err.message.indexOf(msg) >= 0);
-      assert(data === null || data === undefined, 'There should only be an error.');
-      done();
-    } catch (e) {
-      done(e);
+  // this is the queue of chunks that are written to the database
+  // the queue's payload will be an array of documents to be written,
+  // the size of the array will be bufferSize. The variable parallelism
+  // determines how many HTTP requests will occur at any one time.
+  const q = async.queue(function(payload, cb) {
+    // if we are restoring known revisions, we need to supply new_edits=false
+    if (payload.docs && payload.docs[0] && payload.docs[0]._rev) {
+      payload.new_edits = false;
+      debug('Using new_edits false mode.');
     }
-  };
-}
 
-function assertNoError(done) {
-  return function(err, data) {
-    try {
-      assert(err === null, 'There should be no error message.');
-      done();
-    } catch (e) {
-      done(e);
-    }
-  };
-}
-
-describe('#unit Validate arguments', function() {
-  it('returns error for invalid URL type', function(done) {
-    validateArgs(true, {}, assertErrorMessage('Invalid URL, must be type string', done));
-  });
-  it('returns no error for valid URL type', function(done) {
-    validateArgs(goodUrl, {}, assertNoError(done));
-  });
-  it('returns error for invalid (no host) URL', function(done) {
-    validateArgs('http://', {}, assertErrorMessage('Invalid URL', done));
-  });
-  it('returns error for invalid (no protocol) URL', function(done) {
-    validateArgs('invalid', {}, assertErrorMessage('Invalid URL', done));
-  });
-  it('returns error for invalid (wrong protocol) URL', function(done) {
-    validateArgs('ftp://invalid.example.com', {}, assertErrorMessage('Invalid URL protocol.', done));
-  });
-  it('returns error for invalid (no path) URL', function(done) {
-    validateArgs('https://invalid.example.com', {}, assertErrorMessage('Invalid URL, missing path element (no database).', done));
-  });
-  it('returns error for invalid (no protocol, no host) URL', function(done) {
-    validateArgs('invalid', {}, assertErrorMessage('Invalid URL', done));
-  });
-  it('returns error for invalid buffer size type', function(done) {
-    validateArgs(goodUrl, { bufferSize: '123' }, assertErrorMessage('Invalid buffer size option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns error for zero buffer size', function(done) {
-    validateArgs(goodUrl, { bufferSize: 0 }, assertErrorMessage('Invalid buffer size option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns error for float buffer size', function(done) {
-    validateArgs(goodUrl, { bufferSize: 1.23 }, assertErrorMessage('Invalid buffer size option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns no error for valid buffer size type', function(done) {
-    validateArgs(goodUrl, { bufferSize: 123 }, assertNoError(done));
-  });
-  it('returns error for invalid log type', function(done) {
-    validateArgs(goodUrl, { log: true }, assertErrorMessage('Invalid log option, must be type string', done));
-  });
-  it('returns no error for valid log type', function(done) {
-    validateArgs(goodUrl, { log: 'log.txt' }, assertNoError(done));
-  });
-  it('returns error for invalid mode type', function(done) {
-    validateArgs(goodUrl, { mode: true }, assertErrorMessage('Invalid mode option, must be either "full" or "shallow"', done));
-  });
-  it('returns error for invalid mode string', function(done) {
-    validateArgs(goodUrl, { mode: 'foobar' }, assertErrorMessage('Invalid mode option, must be either "full" or "shallow"', done));
-  });
-  it('returns no error for valid mode type', function(done) {
-    validateArgs(goodUrl, { mode: 'full' }, assertNoError(done));
-  });
-  it('returns error for invalid output type', function(done) {
-    validateArgs(goodUrl, { output: true }, assertErrorMessage('Invalid output option, must be type string', done));
-  });
-  it('returns no error for valid output type', function(done) {
-    validateArgs(goodUrl, { output: 'output.txt' }, assertNoError(done));
-  });
-  it('returns error for invalid parallelism type', function(done) {
-    validateArgs(goodUrl, { parallelism: '123' }, assertErrorMessage('Invalid parallelism option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns error for zero parallelism', function(done) {
-    validateArgs(goodUrl, { parallelism: 0 }, assertErrorMessage('Invalid parallelism option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns error for float parallelism', function(done) {
-    validateArgs(goodUrl, { parallelism: 1.23 }, assertErrorMessage('Invalid parallelism option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns no error for valid parallelism type', function(done) {
-    validateArgs(goodUrl, { parallelism: 123 }, assertNoError(done));
-  });
-  it('returns error for invalid request timeout type', function(done) {
-    validateArgs(goodUrl, { requestTimeout: '123' }, assertErrorMessage('Invalid request timeout option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns error for zero request timeout', function(done) {
-    validateArgs(goodUrl, { requestTimeout: 0 }, assertErrorMessage('Invalid request timeout option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns error for float request timout', function(done) {
-    validateArgs(goodUrl, { requestTimeout: 1.23 }, assertErrorMessage('Invalid request timeout option, must be a positive integer in the range (0, MAX_SAFE_INTEGER]', done));
-  });
-  it('returns no error for valid request timeout type', function(done) {
-    validateArgs(goodUrl, { requestTimeout: 123 }, assertNoError(done));
-  });
-  it('returns error for invalid resume type', function(done) {
-    validateArgs(goodUrl, { resume: 'true' }, assertErrorMessage('Invalid resume option, must be type boolean', done));
-  });
-  it('returns no error for valid resume type', function(done) {
-    validateArgs(goodUrl, { resume: false }, assertNoError(done));
-  });
-  it('returns error for invalid key type', function(done) {
-    validateArgs(goodUrl, { iamApiKey: true }, assertErrorMessage('Invalid iamApiKey option, must be type string', done));
-  });
-  it('returns error for key and URL credentials supplied', function(done) {
-    validateArgs('https://a:b@example.com/db', { iamApiKey: 'abc123' }, assertErrorMessage('URL user information must not be supplied when using IAM API key.', done));
-  });
-  it('warns for log arg in shallow mode', function(done) {
-    captureStderr();
-    try {
-      validateArgs(goodUrl, { mode: 'shallow', log: 'test' }, function(err, data) {
-        assert.ok(err);
-        assert.ok(!data);
-        assert(capturedStderr.indexOf('The options "log" and "resume" are invalid when using shallow mode.') > -1, 'Log warning message was not present');
+    if (!didError) {
+      db.service.postBulkDocs({
+        db: db.db,
+        bulkDocs: payload
+      }).then(response => {
+        if (!response.result || (payload.new_edits === false && response.result.length > 0)) {
+          throw new Error(`Error writing batch with new_edits:${payload.new_edits !== false}` +
+            ` and ${response.result ? response.result.length : 'unavailable'} items`);
+        }
+        written += payload.docs.length;
+        writer.emit('restored', { documents: payload.docs.length, total: written });
+        cb();
+      }).catch(err => {
+        err = error.convertResponseError(err);
+        debug(`Error writing docs ${err.name} ${err.message}`);
+        cb(err, payload);
       });
-      done();
-    } catch (e) {
-      done(e);
-    } finally {
-      releaseStderr();
     }
-  });
-  it('warns for resume arg in shallow mode', function(done) {
-    captureStderr();
-    try {
-      validateArgs(goodUrl, { mode: 'shallow', log: 'test', resume: true }, function(err, data) {
-        assert.ok(err);
-        assert.ok(!data);
-        assert(capturedStderr.indexOf('The options "log" and "resume" are invalid when using shallow mode.') > -1, 'Log warning message was not present');
-      });
-      done();
-    } catch (e) {
-      done(e);
-    } finally {
-      releaseStderr();
+  }, parallelism);
+
+  let didError = false;
+
+  // write the contents of the buffer to CouchDB in blocks of bufferSize
+  function processBuffer(flush, callback) {
+    function taskCallback(err, payload) {
+      if (err && !didError) {
+        debug(`Queue task failed with error ${err.name}`);
+        didError = true;
+        q.kill();
+        writer.emit('error', err);
+      }
     }
-  });
-  it('warns for parallism arg in shallow mode', function(done) {
-    captureStderr();
-    try {
-      validateArgs(goodUrl, { mode: 'shallow', parallelsim: 10 }, function(err, data) {
-        assert.ok(err);
-        assert.ok(!data);
-        assert(capturedStderr.indexOf('The option "parallelism" has no effect when using shallow mode.') > -1, 'Log warning message was not present');
-      });
+
+    if (flush || buffer.length >= bufferSize) {
+      // work through the buffer to break off bufferSize chunks
+      // and feed the chunks to the queue
+      do {
+        // split the buffer into bufferSize chunks
+        const toSend = buffer.splice(0, bufferSize);
+
+        // and add the chunk to the queue
+        debug(`Adding ${toSend.length} to the write queue.`);
+        q.push({ docs: toSend }, taskCallback);
+      } while (buffer.length >= bufferSize);
+
+      // send any leftover documents to the queue
+      if (flush && buffer.length > 0) {
+        debug(`Adding remaining ${buffer.length} to the write queue.`);
+        q.push({ docs: buffer }, taskCallback);
+      }
+
+      // wait until the queue size falls to a reasonable level
+      async.until(
+        // wait until the queue length drops to twice the paralellism
+        // or until empty on the last write
+        function(callback) {
+          // if we encountered an error, stop this until loop
+          if (didError) {
+            return callback(null, true);
+          }
+          if (flush) {
+            callback(null, q.idle() && q.length() === 0);
+          } else {
+            callback(null, q.length() <= parallelism * 2);
+          }
+        },
+        function(cb) {
+          setTimeout(cb, 20);
+        },
+
+        function() {
+          if (flush && !didError) {
+            writer.emit('finished', { total: written });
+          }
+          // callback when we're happy with the queue size
+          callback();
+        });
+    } else {
+      callback();
+    }
+  }
+
+  // take an object
+  writer._transform = function(obj, encoding, done) {
+    // each obj that arrives here is a line from the backup file
+    // it should contain an array of objects. The length of the array
+    // depends on the bufferSize at backup time.
+    linenumber++;
+    if (!didError && obj !== '') {
+      // see if it parses as JSON
+      try {
+        const arr = JSON.parse(obj);
+
+        // if it's an array with a length
+        if (typeof arr === 'object' && arr.length > 0) {
+          // push each document into a buffer
+          buffer = buffer.concat(arr);
+
+          // pause the stream
+          // it's likely that the speed with which data can be read from disk
+          // may exceed the rate it can be written to CouchDB. To prevent
+          // the whole file being buffered in memory, we pause the stream here.
+          // it is resumed, when processBuffer calls back and we call done()
+          this.pause();
+
+          // break the buffer in to bufferSize chunks to be written to the database
+          processBuffer(false, done);
+        } else {
+          ee.emit('error', new error.BackupError('BackupFileJsonError', `Error on line ${linenumber} of backup file - not an array`));
+          done();
+        }
+      } catch (e) {
+        ee.emit('error', new error.BackupError('BackupFileJsonError', `Error on line ${linenumber} of backup file - cannot parse as JSON`));
+        // Could be an incomplete write that was subsequently resumed
+        done();
+      }
+    } else {
       done();
-    } catch (e) {
-      done(e);
-    } finally {
-      releaseStderr();
     }
-  });
-});
+  };
+
+  // called when we need to flush everything
+  writer._flush = function(done) {
+    processBuffer(true, done);
+  };
+  return writer;
+};