# UIDs are preserved across serializations

**URL:** <https://forum.sensmetry.com/t/uids-are-preserved-across-serializations/275>\
**Category:** Discussion\
**Tags:** automator, sysmlv2\
**Created:** [October 28, 2025, 1:24am UTC](https://forum.sensmetry.com/t/uids-are-preserved-across-serializations/275 "2025-10-28T01:24:56Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Robert\_Karban](https://yyz2.discourse-cdn.com/flex008/user_avatar/forum.sensmetry.com/robert_karban/32/119_2.png) [@Robert\_Karban](https://forum.sensmetry.com/u/Robert_Karban)\
**Post date:** [October 28, 2025, 1:24am UTC](https://forum.sensmetry.com/t/uids-are-preserved-across-serializations/275/1 "2025-10-28T01:24:56Z")

</div>

Hi,

When parsing/serializing the same model twice in jupyter (with restart kernel in between),  
the nodes get (mostly) the same UUIDs every time, except root namespaces, expressions and literalIntegers. They get new IDs.  
But AttributeUsages have the same ID and then point to all of the new literals and have multiple relationships.  
That messes up visualizations like TomSawyer.

 ![image](https://canada1.discourse-cdn.com/flex008/uploads/syside/original/1X/11f2b6df1c4b2db5d6ebb6ed4e87e9b8ac2d15e9.jpeg)

I thought that every serialization creates new UUIDs every time.

Can you shed some light on this?  
Thanks!  
cc @Dat_Nguyen

---

<div class="post-metadata">

**Author:** ![Daumantas](https://yyz2.discourse-cdn.com/flex008/user_avatar/forum.sensmetry.com/daumantas/32/19_2.png) [@Daumantas](https://forum.sensmetry.com/u/Daumantas)\
**Post date:** [October 28, 2025, 7:21am UTC](https://forum.sensmetry.com/t/uids-are-preserved-across-serializations/275/2 "2025-10-28T07:21:11Z")

</div>

This happens because we do not yet compute `paths` for anonymous elements which in turn creates random UUIDs. We will fix this in a future release together an efficient implementation for serialization to not negatively affect performance.