Workshop Export limitations

Ask the Community
1

Im trying to find the limitations of all possible ways to export data-

Object Table Constraint Default(No filters) Filtered(Data) Filtered(Properties) Filtered(Both)
Rightclick
CSV - 10000 Filtered(Data)<=10000 Filtered(Properties)<=10000 Filtered(Both-10000)
Excel UTC (timezone) 200000 Filtered(Data)<=200000 Filtered(Properties)<=200000 Filtered(Both-200000)
Other 10000 Filtered(Data)<=10000 Filtered(Properties)<=10000 Filtered(Both-10000)
Custom
Function Backed Export (using a “Button” onclick)
CSV - 10k
Excel - 10k

But when writing custom “function backed export”, I find it inconsistent. Sometimes Im able to export >10k and based on number of properties this may vary.

Currently we try to aggregate by groups and not be too granular with the export data to stay within 10k limit.
So, it would be great to know if there is a walkaround to this hard stop.

2

Hey,

Why does it vary below 10k? The function consumes memory/time per property: timeout kicks in before hitting the limit.

Workarounds to go beyond 10k:

  1. Native Workshop export (non function-backed) → Excel supports up to 200k rows

  2. Materialize as dataset → export directly → no row limit (Palantir’s official recommendation)

  3. PySpark transform → dataset → export → no row limit, preserves sort order [Palantir Developer Community]

  4. Contour → up to 100k rows

(Tip: click the arrow to expand details)

If this helped, please mark it as the solution

Ty,

Felipe Montes
Development Advisor
Follow me @ linkedin.com/in/felipemontesl/ or GitHub: Brazillianerd

3

Hi @montes, thanks for your solution.
We have a customised way of doing this download. We let the user pick the properties from the object type, Also we allow them to filter the data based on a function. So on the filtered data and customised column selection; we want to download exact data that is finaly visible in the object table. Currently, I was able to do XLSX export. But with limited number of objects.
As the bottleneck/threshold/limit is “Workshop timeout” which is by default ~

Limit Value Bottleneck
Workshop execution timeout ~7 seconds Object loading (rows.all() ≈ 0.67ms/row)
Workshop response payload ~8 MB base64 XLSX file size (≈ content-dependent bytes/cell)

Main problem :

Function → rows.all() → full objects in memory → XLSX → base64 string → Workshop HTTP response
↑ bottleneck 1 ↑ bottleneck 2 ↑ bottleneck 3
(timeout) (CPU/memory) (8MB payload)

4

Hey, @Benutzer7

The three bottlenecks you identified are correct. Since you already have custom column selection + function-based filtering, here’s what to do:

Short term, squeeze more rows out of the current approach:

  • After filtering, project down to only the selected properties before rows.all(). Loading full objects and trimming after is what kills you on memory/time.
  • Avoid rows.all() if you can — use rows.take(n) in chunks if the function API allows it.

Real fix — go async:
Instead of Function → XLSX → Workshop response, do:

  1. Action triggers a Transform/pipeline with the filter params + selected columns
  2. Pipeline writes the XLSX to a file in Foundry
  3. Workshop polls or shows a “ready” state, then serves a download link

No timeout, no 8MB cap, no row limit. The user waits a few extra seconds but gets everything.

The synchronous path (Function → base64 → Workshop) will always have a ceiling. The only way around all three bottlenecks at once is moving the heavy lifting out of the Workshop request cycle.

Felipe Montes
Development Advisor
Follow me @ linkedin.com/in/felipemontesl/ or GitHub: Brazillianerd