# Writing parents and children docs **URL:** \ **Category:** Elasticsearch\ **Created:** [September 27, 2012, 9:30pm UTC](https://discuss.elastic.co/t/writing-parents-and-children-docs/9178 "2012-09-27T21:30:04Z")\ **Posts on this page:** 4\ **Page:** 1
**Author:** ![phill](https://avatars.discourse-cdn.com/v4/letter/p/779978/32.png) [@phill](https://discuss.elastic.co/u/phill)\ **Post date:** [September 27, 2012, 9:30pm UTC](https://discuss.elastic.co/t/writing-parents-and-children-docs/9178/1 "2012-09-27T21:30:04Z")
I am converting a table of one type of records into two ES doc types. The Parent doc has a 1:n relationship to the children. My Sequence given a sourceRecord is String esDocId = findESParent(sourceRecord); if ( esDocId == null) { esDocId = insertESParent(sourceRecord); } insertChild(sourceRecord, esDocId); When inserting do I have to do a refresh before this sequence to make sure that the parent types are findable, so I don't create 2 of the same parent? If I knew I recently inserted the parent, could I insert the child without worrying if a get/search would find it? My feeling is that many 'families' of parent and children often (but not always) appear together in my input stream, so If I recorded the last n parent IDs I was working with in a ring buffer, then I could save many refreshes when I already knew of an existing parent. ``` String esDocId = findESParentInCache(sourceRecord); // try ``` the cache that remembers 'recently' seen parents. if ( esDocId == null ) { refresh.. esDocId = getESParentByID(sourceRecord); // hit ES to find see if it is an existing parent. if ( esDocId == null) { esDocId = insertESParent(sourceRecord); // create a new parent } addToCache(esDocId); // remember it in case it comes soon } insertChild(sourceRecord, esDocId); Comments? -Paul -- ---
**Author:** ![Lukas\_Vlcek1](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/lukas_vlcek1/32/819_2.png) [@Lukas\_Vlcek1](https://discuss.elastic.co/u/Lukas_Vlcek1)\ **Post date:** [September 28, 2012, 5:56am UTC](https://discuss.elastic.co/t/writing-parents-and-children-docs/9178/2 "2012-09-28T05:56:53Z")
I have not tested it myself but I think you should be able to use real time get for parents no matter the refresh and having parent id is all you need to index child documents. It would be interesting to know if this strategy works. Lukas On Thursday, September 27, 2012, P. Hill wrote: > I am converting a table of one type of records into two ES doc types. > The Parent doc has a 1:n relationship to the children. > My Sequence given a sourceRecord is > String esDocId = findESParent(sourceRecord); > if ( esDocId == null) { > esDocId = insertESParent(sourceRecord); > } > insertChild(sourceRecord, esDocId); > > When inserting do I have to do a refresh before this sequence to make sure > that the parent types are findable, so I don't create 2 of the same parent? > If I knew I recently inserted the parent, could I insert the child without > worrying if a get/search would find it? > > My feeling is that many 'families' of parent and children often (but not > always) appear together in my input stream, so If I recorded the last n > parent IDs I was working with in a ring buffer, then I could save many > refreshes when I already knew of an existing parent. > > ``` > String esDocId = findESParentInCache(**sourceRecord); // try > > ``` > > the cache that remembers 'recently' seen parents. > if ( esDocId == null ) { > refresh.. > esDocId = getESParentByID(sourceRecord); // hit ES to > find see if it is an existing parent. > if ( esDocId == null) { > esDocId = insertESParent(sourceRecord); // create a > new parent > } > addToCache(esDocId); // remember it in case it comes > soon > } > insertChild(sourceRecord, esDocId); > > Comments? > -Paul > > -- -- ---
**Author:** ![phill](https://avatars.discourse-cdn.com/v4/letter/p/779978/32.png) [@phill](https://discuss.elastic.co/u/phill)\ **Post date:** [September 28, 2012, 4:05pm UTC](https://discuss.elastic.co/t/writing-parents-and-children-docs/9178/3 "2012-09-28T16:05:19Z")
On 9/27/2012 10:56 PM, Lukáš Vlček wrote: > I have not tested it myself but I think you should be able to use real > time get for parents no matter the refresh and having parent id is all > you need to index child documents. It would be interesting to know if > this strategy works. Oh I see that the doc says "By default, the get API is realtime, and is not affected by the refresh rate of the index (when data will become visible for search)." So that matches your statement. Get is different than search, thus I need not figure out when to refresh as long as my findESParent(sourceRecord) can use a get to check the index. Luckily the sourceRecord does contain my choice for \_id. Given that I already know the ID, I was thinking of just asynchronously throwing both the parent (when needed) and the child at the index without waiting for either and then cleaning up after myself only if an insert (future) fails. I wonder what a river does when inserting one or two documents and a particular insert fails. Anyone have any comments about "rollBack" of a set of inserts? -Paul > On Thursday, September 27, 2012, P. Hill wrote: > > ``` > I am converting a table of one type of records into two ES doc types. > The Parent doc has a 1:n relationship to the children. > My Sequence given a sourceRecord is > String esDocId = findESParent(sourceRecord); > if ( esDocId == null) { > esDocId = insertESParent(sourceRecord); > } > insertChild(sourceRecord, esDocId); > > ``` -- ---
**Author:** ![system](https://us1.discourse-cdn.com/elastic/original/3X/1/a/1ac57faf039f6b580b3f104ef42a2a89e41014de.png) [@system](https://discuss.elastic.co/u/system)\ **Post date:** [July 6, 2017, 3:10am UTC](https://discuss.elastic.co/t/writing-parents-and-children-docs/9178/4 "2017-07-06T03:10:50Z")