while (true) {
- Thread starter hanser
- Start date
- Status
Does anyone here work somewhere with a "warnings as errors" policy? I like the idea behind it, but it is awfully annoying when developing new code to deal with things like unreferenced variables when you just haven't filled in that section of code yet.
Are there any practices you use to make these cases less annoying?
Are there any practices you use to make these cases less annoying?
I'm sure everyone does things differently. Though it can't be that uncommon to do something like create a function and leave it blank while implementing other stuff. Yes, you can leave out the variable names for the parameters or put in a harmless reference. But that's kind of obnoxious.
I believe you would still get unused variable warnings if all you do is throw an exception...
LordHunter317":3yp5hmao said:
This brings up an interesting tangential point. If I'm writing a large amount of new code, I've always had the habit of compiling it at incremental steps. This seems like the only sane thing to do to me. But perhaps that's a stylistic difference, and why some people don't have these sorts of issues.
I wonder how many people have the pleasure of using a modern DVCS at work. I imagine most of us are not so lucky.
Oh certainly it's worth it. But this makes a whole lot of assumptions that really can't be applied universally. Plenty of people are not in a position in their life where they can just change jobs on a whim. Nor is it really reasonable to think someone would change jobs because of version control.
And, though it may be hard to understand for some people, not everyone is willing or able to take a leadership role. It really shouldn't be a negative mark on a person if they're unwilling to be a flaming extrovert and shove their vision down people's throats.
And, though it may be hard to understand for some people, not everyone is willing or able to take a leadership role. It really shouldn't be a negative mark on a person if they're unwilling to be a flaming extrovert and shove their vision down people's throats.
Blacken00100":2tubfgys said:
This still makes rather sweeping generalizations, though. I've known lots of awesome programmers with great lives and families in "flyover country". Why abandon all that because of a few silly suboptimal things at a company?
Yep, until you have used CVS, SourceSafe, and Alienbrain, quit complaining about SVN 
Another thing to keep in mind about DVCS is that there are definitely industries that it isn't very practical for. For example, I worked in the game industry for 7 years or so, and a single snapshot of the repository for newer games can be 100+GB, since all the art assets need to be versioned with the source. Though this situation seems to be improving with some DVCS giving more thought to shallow checkouts of binaries.
While we're on the topic of stupid stuff everyone always sticks at the top of source files, can anyone give a good reason for putting the filename at the top of each file? What scenario could there be where you don't know the name of the file you're looking at?
As one of the shrinking group of people who actually enjoy debugging memory/low level issues, I say good riddance! I make good money due to my willingness to dive into junk like that.
I'm looking for the proper term for a UI widget. What would you call the "timeline" in media players that you can click around on to move between various points of a video? The best thing I can think of is "scrubber" and "timeline", but my googling is failing me.
To be more specific, I'm working on implementing something like this, except where data is still coming in as the user is scrolling through existing data. The closest analogy is probably time shifting a live TV program. There are some implementation details to work out, especially regarding positioning of the current time indicator when it's being dragged as more data is streaming in.
To be more specific, I'm working on implementing something like this, except where data is still coming in as the user is scrolling through existing data. The closest analogy is probably time shifting a live TV program. There are some implementation details to work out, especially regarding positioning of the current time indicator when it's being dragged as more data is streaming in.
Basically, the last couple of posts have hit on my dilemma, which is why I'm trying to do some more research on how other people have solved this. It's conceivable that a large amount of data could come in while the user is scrubbing around.
I think I would need to treat it as if the new data doesn't exist from a scaling perspective while the user is actively dragging around. Then re-scale when they release. Though that would then cause a jump in positioning out from under the cursor on release. Which is odd.
Or come up with a completely different way of cycling through the data, which escapes me.
I think I would need to treat it as if the new data doesn't exist from a scaling perspective while the user is actively dragging around. Then re-scale when they release. Though that would then cause a jump in positioning out from under the cursor on release. Which is odd.
Or come up with a completely different way of cycling through the data, which escapes me.
I suppose I could go into a little more detail. I'm working on a memory profiler with a heavy emphasis on visual representation of memory layout over time. So the "data" here is largely information about memory allocations. The use case is someone seeing something "interesting" in the profile as the target application is running, and wanting to check it out while continuing to collect data.
It may be that most people just want to collect a profile run and analyze it later. But I've built the tool from the start to be able to deal with the data in real time, which has required being very careful about time complexity of its various algorithms. I think this is a good selling point.
The eventual plan (which - somewhat frighteningly - is quickly approaching v0.1) is to sell it via a MicroISV-type setup. Perhaps I'll post more sometime post-launch to share about its success/failure.
It may be that most people just want to collect a profile run and analyze it later. But I've built the tool from the start to be able to deal with the data in real time, which has required being very careful about time complexity of its various algorithms. I think this is a good selling point.
The eventual plan (which - somewhat frighteningly - is quickly approaching v0.1) is to sell it via a MicroISV-type setup. Perhaps I'll post more sometime post-launch to share about its success/failure.
[url=http://meincmagazine.com/civis/viewtopic.php?p=24889153#p24889153:3iu617ke said:
This is an awesome idea. I hope more languages adopt it!
Very frustrating when I encounter someone who doesn't "get" DVCS and thinks it's stupid.
I've actually known very few people who prefer something other than VS, and most of my career has been in C++. Though across multiple companies, everyone insists on having Visual Assist. There have been a couple times I had license issues and had to briefly work without it. My productivity slows to a crawl.
Usually, the source code is available for things like delete. Check it out. It just calls free under the hood. Even if you allocated a class with malloc and then called delete, it would, in theory, be fine. Unless your destructor does stuff that expects state setup in the constructor.
I agree with that structure. Boolean tests should be either positive or negative (positive, preferally) across the codebase, when possible. Otherwise it's too easy to get confused.
I would like to make a confession. I suck at trees. If a tree problem comes up in an interview, I might as well cross that job off the list.
Sure. But interview questions tend to focus more on gotchas and unusual operations on trees.
Here's one I was recently asked. Apparently, I actually did better than I thought, as I'm going to the next round. The problem is really quite simple, but there's a linear-time solution that just wasn't popping into my head.
Given a tree where each node only contains a pointer to its parent, find the first common ancestor for any two arbitrary nodes.
Part of my hang up was that I was trying to think about lots of complicated trees, when the problem actually boils down to two lists converging. The main issue is dealing with finding the point of convergence without storing extra information or doing an n^2-type loop. If the two given nodes are at different heights in the tree, you can't simply walk up the tree and expect to get a collision. Visualized:
Once you know the obvious solution, it is obvious. You need to measure the height of both nodes by walking all the way up for both of them. Then you "trim up" the longer tail so its height matches the shorter one. Then walking them up side-by-side will produce a guaranteed collision.
Given a tree where each node only contains a pointer to its parent, find the first common ancestor for any two arbitrary nodes.
Part of my hang up was that I was trying to think about lots of complicated trees, when the problem actually boils down to two lists converging. The main issue is dealing with finding the point of convergence without storing extra information or doing an n^2-type loop. If the two given nodes are at different heights in the tree, you can't simply walk up the tree and expect to get a collision. Visualized:
Code:
1
2
3
4 5
6 7
8Once you know the obvious solution, it is obvious. You need to measure the height of both nodes by walking all the way up for both of them. Then you "trim up" the longer tail so its height matches the shorter one. Then walking them up side-by-side will produce a guaranteed collision.
Just with a cursory scan of your solution, that looks right. The actual code, unfortunately, is the easy part. It's lucking out and getting that eureka moment that's the issue for me. Especially in the time constraints and stress of an interview situation.
I once built a dynamic inheritance system in C# via reflection. It's not my proudest moment.
I didn't like WPF when I was first learning it, but it has really grown on me. Once you get past the steep learning curve, it's pretty rad.
Swagger is wonderful, but it takes like a day or two to figure out the basics. Seems silly to require it as a pre-existing skill.
I spent years trying to figure out why Windows 10 was snappy on some machines and hideously slow on others. The miracle cure ended up being the first fix on this page about disabling Cortana: https://windowsreport.com/file-explorer ... indows-10/
Sounds crazy, but I've done that on multiple machines now and it makes a massive difference.
Sounds crazy, but I've done that on multiple machines now and it makes a massive difference.
The fact that they say this is proof that their interview process is broken.
Count() operates on an IEnumerable and literally counts it, vs Count already has the correct value. So O vs. O(1).
vs. O(1).Far too easy to introduce bugs without liberal parens. This is one of the few hills I'm willing to die on, as I've fixed too many said bugs
I somehow have managed to have my own codebase with no one else working on it going on 3 years now at a Big Tech company. No code reviews for me!
- Status