Today I got turned down by this interview question, which is actually surprisingly for pre-TMP UGUI text. They wanted as many solutions as possible. I searched around and found that the image has the function alphaHitTestMinimumThreshold, which addresses the click-through problem for Unity UI needs. Meanwhile, GPT is telling me to go character by character in cacheTextGenerator and draw a circle at the right place. This is only an approximation. They didn't like the GPT answer, so I am wondering if there is a better way to do it.
#Click UGUI text (not TMP) with letter "O" in it, detect clicked on ring or empty space.
1 messages · Page 1 of 1 (latest)
no reply 🙁
why did you feed them a GPT answer? Their question was a little confusing, did they mean a UGUI Image with text on it? Because saying UGUI text (not TMP) made me think of the ancient pre-TMP UGUI text that nobody uses
I'm pretty sure they wanted the text component, not as an image. I was basically answering their previous questions with pre-TMP UGUI, which I had little practice with. I am very confused about why they ask for such things, so I thought there must be something obvious I missed.
And they wanted multiple solutions for each question, So i fed them one of the GPT solutions and some of my "cheating" ways, which none of them felt right.
My cheating way was adding 2 colliders for inner and outter area. But they don't fit perfectly with the "O"
What was the context the questions were asked in? If this was a discussion, this would've been a great opportunity to point out that they're solving the wrong problem if this is important and pushed back on legacy GUI
HR was there to answer my question, i sort of asked "can I assume you are asking about TMP instead of non-TMP text" , they replied "nope"
was it HR or was it an engineer? because now my opinion has changed, HR reads from a script and knows nothing
Goes like this
"Hi, Can I assume the text in the 2 quetsion are TMP?"
"please wait, let me check with the team. ..... their reply is 'U can not.'"
"OK"
you might've gotten hosed by HR asking questions they don't understand

did you push back on that design decision? My answers would've been, "why aren't we using TMP? I would convince the team to use TMP instead of this legacy component" and "okay, if this feature is important, render it to an Image instead if it can't be TMP"
the chatgpt answer breaks down terribly at even the slightest breeze. What about different fonts? Font styles? What if the real question is to detect clicks on the exact border of text with any char?
sorry dude, super unlucky to get turned down by a dumb brainteaser asked by an HR drone
There may be a possibility that they wanted to check if you would try to bullshit them by trying to give a solution for an unrealistic problem. 🤔
GPT answer goes into font vertices to check for current font "O"'s width, and search for letter, from my tests, it is actually legit. But, no, I didn't push back on the decision, because it was the first two questions I wasn't expecting this kind of difficulty. I'm not sure, but I want to learn if there is a legitimate way to check text-body vs. text void spaces in legacy UGUI text.
I BSed a little
When I think about it, giving such questions is actually pretty decent way to detect if someone is using ChatGPT. The AI tends to bullshit a lot.
They told me there would be a card game design demo test later, so I was not expecting a legacy-text UGUI question. Bombed my first written interview.
I have used UGUI for many years in released games and I mostly use TMP only. I legit had no idea this event hit threshold existed on Image but perhaps this question was testing general problem solving? (I think this because TMP has been the replacement for legacy text for a while now)
The docs description for "alphaHitTestMinimumThreshold" informs us that the texture is "sampled" cpu side to determine alpha and this can easily be done ourselves with any mesh if we know the texture and UV of the "interaction".
I think if I saw this question myself id be questioning when Id ever do this but hey go interviews can be like that 🤷♂️
An alternate answer they probably did want was probably some distance check to determine if we were inside the area of a ~~circle ~~ring in some rect which is pretty basic stuff.
that was my initial thought: maybe they wanted to see if you could recognize the problem itself was dumb. But it wasn't an engineer asking the questions but some HR person instead
HR uses ai to produce bullshit questions
candidate uses ai to produce semi bullshit answer
job market is great rn oh boy 😐 /s