Known limitations
-
Processes which have loops and those loops are sharing the same cells may result in a process deadlock (infinitely-running process). For now this is an unsupported pattern. If this happens, cancel the process using the process menu in the top right hand corner. We may detect those patterns and provide better feedback in the future.
-
Publishing very large datasets (e.g. hundreds of MB) using
datacards.variable.publish()can lead to performance bottlenecks, since variables are synced with frontends on demand. Therefore we limit to 500Mib for publishing variables. We recommennd using the filesystem (e.g. with a sqlite file) or an external database for large amounts of data, and publishing just proxy variables to describe your data flow between notebooks. -
The more notebooks your project uses, the more RAM will be consumed. For example, if your project has 4GB of RAM, it will be possible to create around 20 notebooks before the RAM is full. If you need more compute resources, please contact us. For more on the tradeoffs and some design patters, see the best practices 
-
Currently, the
keyproperty of DataCards variables must conform to the Python variable naming rules . Publishing variables with non-conformingkeynames is possible, but they will throw an error when they are consumed. -
We do no code analysis of cards and variables. You need delete the cell (not the code in the cell) to remove the card or variable.
Supported browsers and hardware
- For the best experience, we recommend using one of the three most recent stable releases of Google Chrome . Other browsers (e.g. Safari) may work, but we target our bug-fixing efforts on Chrome.
- DataCards is conceived and optimized for desktop devices. Mobile devices are not yet supported.