When you import certain versoin of docx docment into Google Doc, there could be a lot of bookmarks, making the UI look a rather messy. See some related but unresolved discussions a few year back.
Now that we have the quick solution.
Solution
Open "Extensions --> App Script". Copy and paste the below. Then execute the function.
function removeAllBookmarks() {
var doc = DocumentApp.getActiveDocument();
var bookmarks = doc.getBookmarks();
for (var i = bookmarks.length - 1; i >= 0; i--) {
var bookmark = bookmarks[i];
var element = bookmark.getPosition().getElement();
console.log(bookmark);
if(element){
try {
bookmark.remove();
} catch (e) {
console.log(e)
}
}
}
}
The code is generated by AI, with a bit post editing.
The exploration
AI's first cut
function removeAllBookmarks() {
var doc = DocumentApp.getActiveDocument();
var body = doc.getBody();
var bookmarks = doc.getBookmarks();
for (var i = bookmarks.length - 1; i >= 0; i--) {
var bookmark = bookmarks[i];
bookmark.removeFromDocument();
}
}
There was an error:
TypeError: bookmark.removeFromDocument is not a function
AI's second try
We feedback the error to AI and ask for a revision.
function removeAllBookmarks() {
var doc = DocumentApp.getActiveDocument();
var bookmarks = doc.getBookmarks();
for (var i = bookmarks.length - 1; i >= 0; i--) {
var bookmark = bookmarks[i];
var element = bookmark.getPosition().getElement();
if(element){
bookmark.removeFromDocument();
}
}
}
The trial looks a bit random. It did not analyze the root cause of "removeFromDocument is not a function" is likely a change of funtion name.
Human intervention
We take closer look at the bookmark object:
console.log(bookmark);
Here is the output:
{ toString: [Function],
remove: [Function],
getId: [Function],
getPosition: [Function] }
The solution is to change from removeFromDocument to remove.
Tada! It works.
One more thing
The function worked on most bookmarks, but there were a few instances of the below error:
Exception: Unexpected error while getting the method or property remove on object DocumentApp.Bookmark.
I din't know the root case in the end. Those bookmarks seem invisible and do not interfere with my further editing, so I decide to let them go and use a try-catch to skip those errors.
That completes the solution.
Review
The total time spent on the case is about 5min (excluding writing this blog). Model is Gemini 2.0.
In the past, this kind of occasional demand can drive me crazy. Open Internet search often leads to fragmented discussion, and putting together those materials is often time consuming.
AI's code generation capability gets us much faster on the ground works. Sometimes they have knowledge gap due to outdated documentation / unknown user execution environment. This part of troubleshooting is quick for an experienced programmer.
This case again demonstrates the Geek Efficiency Curve Updated. Human + AI collaboration boost efficiency of tasks that are too large to treat manually, and are too small to invest in solid engineering resources.
Other observations
- LLM+Search is definitely the new paradigm. The output from Perplexity gives a more direct answer. Although the search results include 80% useless / irrelevant content, the LLM is able to quickly "glance through" and extract the most important part.
- In this new paradigm, "main streaming" (of solution/ of opnion) will become an increasing concern. In the above example, the Perplexity answer is mainly based on the gist. The original gist is certainly a very good answer using API (Doc's AppScript automation). Down there you can find another nice hack, that is to simulate user behaviour by clicking on the every "remove" button. Both solutions are great. The API one is more stable and future proof. The user behaviour simulation one is creative and enables us to thinking out of the box!! In the future, we may encounter the cases when API is not supported yet, but we can simluate a user behaviour to complete the tasks.
