eg. 把pmid列由字符型转为整型

1 find + forEach逐条修改

文档较大时会很慢

  1. # 更改全部数据
  2. db.article.find().forEach(function(x) {
  3.     x.pmid = NumberInt(x.pmid);
  4.   db.article.save(x);
  5. })
  7. # 只更改类型为string的数据
  8. db.article.find({pmid: {$type: 2}}).forEach(function(x) {
  9.     x.pmid = NumberInt(x.pmid);
  10.   db.article.save(x);
  11. })

2 bulkWrite批量修改

mongo version >= 2.6 and < 3.2

  1. var col = db.article;
  3. var bulk = col.initializeOrderedBulkOp();
  4. var counter = 0;
  6. col.find({ pmid: { $type: "string" }}).forEach(function(data) {
  7.     var updoc = {
  8.         "$set": {}
  9.     };
  10.     updoc["$set"]["pmid"] = NumberInt(data.pmid);
  11.     // queue the update
  12.     bulk.find({
  13.         "_id": data._id
  14.     }).update(updoc);
  15.     counter++;
  16.     // Drain and re-initialize every 1000 update statements
  17.     if (counter % 1000 == 0) {
  18.         bulk.execute();
  19.         bulk = col.initializeOrderedBulkOp();
  20.     }
  21. });
  23.     // Add the rest in the queue
  24. if (counter % 1000 != 0) bulk.execute();

mongo version >= 3.2

  1. var bulkOps = [];
  3. db.article.find({pmid: {$type: 2}}).forEach(function(x) {
  4.     var new_pmid = new NumberInt(x.pmid);
  5.     bulkOps.push({ 
  6.         "updateOne": { 
  7.             "filter": {"_id": x._id },              
  8.             "update": {"$set": {"pmid": new_pmid}} 
  9.         }         
  10.     });
  11. })
  13. db.article.bulkWrite(bulkOps, {"ordered": true});
  15. # ============================================================
  16. use pubmed;
  18. var counter = 0;
  19. bulk = [];
  21. db.article.find({pmid: {$type: 2}}).forEach(function(x) {
  22.     var new_pmid = new NumberInt(x.pmid);
  23.     bulk.push({
  24.         updateOne: {
  25.             filter: {_id: x._id },
  26.             update: {$set: {pmid: new_pmid}}
  27.         }
  28.     });
  29.     counter++;
  30.     if (counter % 100000 == 0) {
  31.         print(counter, 'records updated');
  32.         db.article.bulkWrite(bulk, {ordered: true});
  33.         bulk = [];
  34.     }
  35. })
  36. if (counter % 100000 != 0) db.article.bulkWrite(bulk, {ordered: true});

3 update方法

Mongo 4.2

  1. db.collection.update(
  2.   { a : { $type: 1 } },
  3.   [{ $set: { a: { $convert: { input: "$a", to: 2 } } } }],
  4.   { multi: true }
  5. )

参考：