is there a function for dump the DOM

[raku-cs]

besides the screenshot function, could I have a function that dumps the whole DOM in a tab?

[DCoomer]

Something in Evaluate might work: https://docs.taiko.dev/api/evaluate/

Why do you need the whole DOM?

[AutoAPEAI]

In certain scenarios, having access to the entire DOM (Document Object Model) becomes crucial for precise and intelligent automation. Consider a case where a web page contains multiple elements with identical or similar text content, and the goal is to interact with a specific, context-dependent element.

The complete DOM empowers us to train models or leverage machine learning techniques to identify and target the most relevant element dynamically. By understanding the entire structure of the page, including relationships between elements, classes, and hierarchies, we can build robust automation scripts that adapt to different contexts.

In essence, the full DOM provides the necessary context for intelligent decision-making, particularly in scenarios where there might be ambiguity or overlap in element identification based solely on visible text content. It enables us to create automation that goes beyond basic text matching and contributes to more sophisticated and adaptable workflows.

While not every automation scenario requires the whole DOM, having access to it becomes advantageous when dealing with complex web applications or situations where precise element identification is paramount.

[DCoomer]

This looks like an automated reply that doesn't add value to the thread.

[raku-cs]

Hi, DCoomer

Thanks for your reply and suggestion. The context of mine is that, we are trying to do the machine learning staff (which is similar to https://osu-nlp-group.github.io/Mind2Web/ ). So, if we want to improve the segmentation accuracy or auto-action prediction accuracy, the more data we have, the better performance we could gain.

Therefore, for the data collection phase, screenshot is not enough for us. Furthermore, the more data we have, the more insights we could have, which could guide users to write more concise taiko test codes. (as have been mentioned by AutoAPEAI above, sometimes we cannot write the correct testing code in taiko, due to the complicated web page layout and contents)

[AutoAPEAI]

This looks like an automated reply that doesn't add value to the thread.

My intention is to provide a comprehensive context of the problem that we're trying to solve here: disambiguate Web elements with similar/same texts. And we are taking the approach of training models with DOM data so that the model will intelligently predict the most relevant Web element for completing a specific automation task.

[zabil]

Taiko doesn't have native support for this. However, Taiko exposes cri (refer https://docs.taiko.dev/api/client/) using which you could execute Chrome Devtools commands and extend this to add the functionality.

[AutoAPEAI]

Thanks for your prompt response! 🙌 I'll dig into the Taiko documentation and explore the cri module to work on extending the functionality. If I run into any roadblocks or have questions along the way, I might reach out again. Appreciate your guidance!

[DCoomer]

https://chromedevtools.github.io/devtools-protocol/tot/DOM/

const dom = await client().send("DOM.getDocument",{pierce:true, depth: -1});

May be useful.

Let us know what you find.

[AutoAPEAI]

Thanks for sharing! I'll explore the DOM using the provided code and let you know what insights I find. Appreciate your input!