Try it
Those horizontal and vertical axis labels are canvas elements, so you’re not gonna be able to extract text from them in any easy way.
I did just spend a lot of time trying to figure this out for you, but it’s pretty opaque and limited
I was going to tag you @navitronic , thought this would be right up your street. I haven’t heard of canvas elements so I’m going to look into that, cheers.
Thanks for looking bud, I appreciate it!
They do have a github if you get bored, heh heh
ha, I will not take that bait…
Out of interest, what are you trying to do?
I am collecting data on trades and as part of it have to go through months and months of data backtesting. I need to (accurately) record the start/end time of the dates of the trades and it would be a hell of a lot easier if I could use shortcut keys instead of having to manually type them into a spreadsheet.
I am using smartsheets for this, which accepts dates typed like ‘08/07/23’ and ‘17:00’ but I am used to working with airtable where you don’t need the '/'s or ‘:’ in the time. It would be much easier if I could just ALT+C, copy date in correct format and alt tab straight into the worksheet. But I am dangerously close to spending way too much time wrestling with this problem rather than doing the actual work.
I don’t know if this is what you’are after but you can definitely use formulas in google sheets to convert dates and times from your raw data.
You could even use automated scripts which fire using cell edit triggers so when you paste your raw data it outputs your cleaned up data somewhere.
This type of thing is what chatGPT is really good for. Give it some data, tell it exactly what you want to do and keep shouting at it until it gets it right.
Sounds like getting the raw data is the trickier part of the equation.
Unfortunately Les converting date/times is waaaay down the line here (and yes I have plenty of experience!). I am struggling to even get the data to convert first.
It always appears a set distance from bottom of screen under mouse cursor, I am thinking of using something like AutoHotKey to inspect that part of the screen, try to detect the text then go from there. Hmmmmm
Anyone a GA4 nerd?
I know of it.
If you’re looking at ecom tracking though I won’t be able to help 
1 Like
Theres one referral thats pulling away from email stats.
Titan mail, but I don’t know why its attributing the conversion to that and not insider/mail (the software we use)
What do you mean pulling away?
As in it should be in a report but it’s not?
Its like splitting it, ah I dunno how to explain.
Show us a screenshot
Do they have an API?
Not one that would help. It’s just cursor position on the screen and as nav said it’s a canvas element and there’s no good way of interrogating it.
I’ve got a support ticketl with tradingview and it’s in their backlog so we’ll see what happens.
As for now I am playing with windows power automate to screencap the date/time, OCR it and spit it out to clipboard. Just a bit longwinded aye?!?
I love this shit.
I’d ocr it too. Screencap it using Selenium, feed than into Gemini or some other model and get that to parse it for you.
You don’t need to do your own OCR. BITD I’d use a pretrained model e.g. GCP vision API for that, but nowadays you just need a wrapper for a foundation model to do the heavy lifting. PM me details if U want, I love working on this sort of stuff.
Don’t think it’s possible as you need to know what to query. Usually it’s fairly simple, a div or span or whatever, but these canvas elements can’t be interrogated (easily). And Power Automate has built in OCR so it’s pretty easy and works perfect. Just having problems clipping the exact bit of the screen as you got to use a powershell script for the screencap. At the moment it’s ignoring the coordinates I’m giving and just outputting a huge transparent rectangle. Also it’s got to be quick, otherwise it’s just easier for me to do it all manually. I’ve already spent far too much time trying to solve the problem but I love shit like this.
You can’t grab the whole screen? If you can get the whole screen, you can just chuck it at an LLM and prompt it to figure out what you want from it. Like you did on here with that screenshot above.
There 100% has to be a way to programmatically grab the whole screen without querying an element. I’m gonna try later
