# Optimizing the writing of large amounts of data in neo4j with apoc Parquet, periodic iterate

**URL:** <https://community.neo4j.com/t/optimizing-the-writing-of-large-amounts-of-data-in-neo4j-with-apoc-parquet-periodic-iterate/65077>\
**Category:** Procedures & APOC\
**Tags:** apoc, performance, cypher\
**Created:** [November 24, 2023, 10:03am UTC](https://community.neo4j.com/t/optimizing-the-writing-of-large-amounts-of-data-in-neo4j-with-apoc-parquet-periodic-iterate/65077 "2023-11-24T10:03:16Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![tlaucournet](https://sea1.discourse-cdn.com/flex021/user_avatar/community.neo4j.com/tlaucournet/32/30113_2.png) [@tlaucournet](https://community.neo4j.com/u/tlaucournet)\
**Post date:** [November 24, 2023, 10:03am UTC](https://community.neo4j.com/t/optimizing-the-writing-of-large-amounts-of-data-in-neo4j-with-apoc-parquet-periodic-iterate/65077/1 "2023-11-24T10:03:16Z")

</div>

Hi,  
I need to import hundreds of millions of nodes and relationships into my database while taking care of duplicate.  
I created indexes for most of my nodes.  
I'm using Using neo4j 5.13 and apoc

This is what look like my cypher query

```auto
            CALL apoc.periodic.iterate('
            CALL apoc.load.parquet("file:///my_file.parquet") YIELD value RETURN value
        ','
            MERGE (n1:Label1 {property: value.property1})
            ON CREATE SET n.property2 = value.property2

            FOREACH (_ IN CASE WHEN value.info1 IS NOT NULL AND value.info1 <> \\'\\' THEN [1] ELSE [] END |
                MERGE (n2:Label2 {info: value.info1})
                MERGE (n)-[:HAS_INFO]->(n2)
            )
            
            FOREACH (_ IN CASE WHEN value.info2 IS NOT NULL AND value.info2 <> \\'\\' THEN [1] ELSE [] END |
                MERGE (n3:Label3 {info: value.info2})
                MERGE (n)-[:HAS_INFO]->(n3)
            )

            WITH value.unknown_values AS uvalues, n
            UNWIND uvalues AS uvalue
            MERGE (n4:Label4{info:uvalue})
            MERGE (n)-[:HAS_UNKNOWN_INFO]->(n4)

            ',{batchSize: 10000, parallel: true}
            )

```

I'd like to know if my way of doing things seems to you to be optimized or not.

Thanks

---

<div class="post-metadata">

**Author:** ![glilienfield](https://sea1.discourse-cdn.com/flex021/user_avatar/community.neo4j.com/glilienfield/32/27534_2.png) [@glilienfield](https://community.neo4j.com/u/glilienfield)\
**Post date:** [November 24, 2023, 10:47am UTC](https://community.neo4j.com/t/optimizing-the-writing-of-large-amounts-of-data-in-neo4j-with-apoc-parquet-periodic-iterate/65077/2 "2023-11-24T10:47:39Z")

</div>

It does look typical. Two comments though. I don’t see where the variable ‘n’ that is referenced throughout the code is defined. The first match binds its result to ‘n1’. Should ‘n1’ be ‘n’ instead?

Second, sometimes running the batches in parallel becomes problematic when the code creates relationships. This is because there could be blocking due to the end nodes in a relationship needing to be locked in order to create the relationship. This would occur if the data is creating multiple relationships to the same node. Just a heads up.

---

<div class="post-metadata">

**Author:** ![tlaucournet](https://sea1.discourse-cdn.com/flex021/user_avatar/community.neo4j.com/tlaucournet/32/30113_2.png) [@tlaucournet](https://community.neo4j.com/u/tlaucournet)\
**Post date:** [November 24, 2023, 11:03am UTC](https://community.neo4j.com/t/optimizing-the-writing-of-large-amounts-of-data-in-neo4j-with-apoc-parquet-periodic-iterate/65077/3 "2023-11-24T11:03:56Z")

</div>

Thanks for your answer.  
Indeed I made a mistake, 'n' should be 'n1'  
Regarding lock between batches i have some issue, I am using the 'retries' parameter of apoc.periodic.iterate that i forgot to add in the example.