Hello thanks for the reply! I have figured it out using this code block

group by function { \
    let taskGroup = (task.file.filenameWithoutExtension || "") + (task.heading ? "#" + task.heading : ""); \
    let totalUrgency = query.allTasks.filter(t => \
        (t.file.filenameWithoutExtension || "") + (t.heading ? "#" + t.heading : "") === taskGroup && !t.isDone) \
        .reduce((sum, t) => sum + (t.urgency || 0), 0); \
    let prioGroups = query.allTasks.filter(t => \
        (t.file.filenameWithoutExtension || "") + (t.heading ? "#" + t.heading : "") === taskGroup && t.happens.moment && !t.done.moment) \
        .map(t => (t.file.filenameWithoutExtension || "") + (t.heading ? "#" + t.heading : "")); \
    let sortedGroups = [...new Set(query.allTasks.filter(t => !t.isDone) \
        .map(t => (t.file.filenameWithoutExtension || "") + (t.heading ? "#" + t.heading : "")))] \
        .map(group => ({ \
            group: group, \
            urgency: query.allTasks.filter(t => \
                (t.file.filenameWithoutExtension || "") + (t.heading ? "#" + t.heading : "") === group && !t.isDone) \
                .reduce((sum, t) => sum + (t.urgency || 0), 0) \
        })) \
        .sort((a, b) => b.urgency - a.urgency); \
    let index = sortedGroups.findIndex(g => g.group === taskGroup); \
    return taskGroup ? '%%' + index + '%% [[' + taskGroup + ']] (Total Urgency: ' + totalUrgency.toFixed(2) + ')' : ''; \
} 

Firstly, because i used backlinks to group so i will use task.file.filenameWithouExtension and task.heading to group the same file and heading tasks, then i calculate total urgency score in each group using task.urgency and then i normalize it to avoid bias with many task but low priority problem, then i sort it using hidden text %%index%%
This code is written by chatgpt because im not familiar with js at all, so it a bit long and i have no idea how to optimise this code, i know urgency score is not always correct, but i only know this way to sort group, if you have any recommendation, i would love to edit it!

Result:

Interesting. Well done for getting something that works for you.

I had a quick look at the code and it is very repetitious.

Maybe you could ask it to try creating a few functions to make the code less repetitious. Or see if it can add some comments in to the code, to explain to you how it is working.

I tried running it on my vault, and the performance was so slow then I ended up having to Force-Quit Obsidian each time.

This code is written by chatgpt because im not familiar with js at all, so it a bit long and i have no idea how to optimise this code, i know urgency score is not always correct, but i only know this way to sort group, if you have any recommendation, i would love to edit it!

I no longer have time to help people with custom searches, unfortunately.

That search is basically using Tasks in a way that was not foreseen, and so it is probably doing a lot of very repetitious calculations against each task that is searched, for every task that is searched....

It is clever, and I would not have thought of doing it, but if the performance is a problem for you, I suggest trying to find a simpler approach.

Hello, i end up kept using chatgpt to reduce redundancy and this is final version:

group by function { \
    let allTasks = query.allTasks.filter(t => !t.isDone); \
    let groupUrgencyMap = new Map(); \
    let groupTaskCountMap = new Map(); \
    allTasks.forEach(t => { \
        let taskGroup = (t.file.filenameWithoutExtension || "") + (t.heading ? "#" + t.heading : ""); \
        groupUrgencyMap.set(taskGroup, (groupUrgencyMap.get(taskGroup) || 0) + (t.urgency || 0)); \
        groupTaskCountMap.set(taskGroup, (groupTaskCountMap.get(taskGroup) || 0) + 1); \
    }); \
    let normalize = (group) => groupUrgencyMap.get(group) / (groupTaskCountMap.get(group) || 1); \
    let sortedGroups = [...groupUrgencyMap.keys()].sort((a, b) => normalize(b) - normalize(a)); \
    let taskGroup = (task.file.filenameWithoutExtension || "") + (task.heading ? "#" + task.heading : ""); \
    let index = sortedGroups.indexOf(taskGroup); \
    return taskGroup ? '%%' + index + '%% [[' + taskGroup + ']] (Normalized Urgency: ' + normalize(taskGroup).toFixed(2) + ')' : ''; \
} \

You can try it to see if performance is improved!

I no longer have time to help people with custom searches, unfortunately.

Thanks for mainterning this wonderful plugin!

Thanks.
You should remove the \ from the end of the last line.

@TamadoIchikai Just to let you know, I am working on a way to greatly speed up searches like yours, that calculate a value for many tasks, and currently have to do that repeatedly...

It will need a new Tasks release, and documentation to be written. When it's done, I'll report back here and give you the new version of your group by function to use.

Hello, thanks for the rapid update of the plugin! To try out your code, i simply create a daily template with templater plugin syntax like this,

• Code 1 📅 <%tp.file.title%>
• Code 2 📅 <%tp.file.title%>
• Code 3 📅 <%tp.file.title%>
• Code 4 📅 <%tp.file.title%>
• Code 5 📅 <%tp.file.title%>
• Code 6 📅 <%tp.file.title%>
• Code 7 📅 <%tp.file.title%>
• Code 8 📅 <%tp.file.title%>
• Code 9 📅 <%tp.file.title%>
• Code 10 📅 <%tp.file.title%>
  .
  .
  .
• Code 97 📅 <%tp.file.title%>
• Code 98 📅 <%tp.file.title%>
• Code 99 📅 <%tp.file.title%>
• Code 100 📅 <%tp.file.title%>
  then combine with calendar plugin, i can create many dummy task for my vault, i chose this way because my vault currently doesn't has enought task to see the different. After that, i have around 600 tasks. This is my laptop specs:
• OS: window 11 build 26100
• Processor: intel(R) core i5-8300H CPU with 2.3 GHz
• RAM: 16gb
• eGPU: GTX 1050
• iGPU: intel(R) UHD 630
  After try out your code and my old code, to search for 600 tasks, i can see that there is about 0.2s faster with your code. Sorry for this benchmark method, i didn't have any idea other than try it the old fashioned way, and i couldn't generate more task since this template ate almost 8gb ram when generating. Nonetheless, i could alsolutely notice the speed differents. Thanks for upgrading 😁

Thank you for those numbers.

The time cost goes up with the square of the numbers of tasks.

So yours, with open 600 tasks, went:

• from 360,000 calculations through the search to:
• to 600 calculations.

Mine - with around 1,900 open tasks - went from around:

• from 3,610,000 calculations
• to 1,900 calculations

Which is a much bigger speed up...

@rhhub - excellent.
When you have solved it, please do add a note in #1939 with the code, and maybe I can use it in the documentation as a motivating example for the to-be-written docs on the query.searchCache facility.

Incidentally, I would encourage you to use const instead of let for safety.
(I didn't include it in the example above, as I wanted it to be easier to spot the minimal changes needed to adopt query.searchCache.

Finished the code. It Groups tasks by tags. Sorts the groups by urgency associated with most urgent task contained within the group. It's discussed here: #1939 (comment)

Same max urgency of groups were being calculated from tasks due in the future. Glad it helped.