kz930 opened a new issue, #7326: URL: https://github.com/apache/texera/issues/7326
### What happened? The operator builds its image with `WordCloud(width=1920, height=1080, stopwords=set(STOPWORDS), max_words=..., background_color='white', include_numbers=True)` and passes no `random_state`, so the library draws its word placement and colours from an unseeded generator. Rendering the same text twice produces two images that do not match byte for byte; passing `random_state=0` makes them identical. Only the arrangement moves. Taking the layout apart across two unseeded renders of the same text, the words and their frequencies are identical — `plan` at 1.0, `bus` at 1.0, `transit` at 0.33 and so on — while every position, orientation and colour differs. Which words appear and how large they are is decided by frequency alone and is not affected by the seed, so nothing about the analysis is at stake here. What is at stake is reproducibility. Re-running a workflow rearranges every word, a saved screenshot stops matching what the operator now produces, and two people opening the same workflow see different pictures of the same data — with no way to recover the earlier rendering, including one the user happened to prefer. Every other visualization in Texera returns the same picture for the same input. Two ways to fix it, and the choice is a product call rather than a technical one: - Pass a fixed `random_state`. Smallest change, makes the operator behave like every other chart. Existing workflows render one last different picture and are stable from then on. - Expose the seed as a `Layout Seed` setting with a fixed default. Same reproducible default, and a user who wants a different arrangement can change the number and can get that arrangement back later, which is not possible today. Expected either way: a given input renders the same image every time. ### How to reproduce? Point a Word Cloud at any text column and run the workflow twice without changing anything. The same words appear at the same sizes, placed at different positions and angles and drawn in different colours. ### Version/Branch 1.3.0-incubating-SNAPSHOT (main) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
