Summary: I am going to have to do a lot more practice, and brush up on some things, especially dynamic programming, if I’m going to do well on medium to hard questions.

The good:

• The practice in the past few weeks has got the idiosyncrasies of Go slices and arrays mostly under my belt. Stuff I was struggling with in the first few exercises is now pretty reflexive. I’m reasonably certain that I can manipulate slices as I need to.
• I really do remember pointer algorithms from my systems programming days. I was, in some cases, finding it easier to implement algorithms using linked lists for variable linear storage than slices. (Slices have improved a lot though, so I’m not reaching for linked lists first.)
• I’m doing okay with recursion now.
• Tree algorithms are not as bad. The medium height-order problem was easy for me to conceptualize; I just had trouble getting the slice of slices to work.
• If the code can use a queue-based algorithm, I have no problem writing it. 01 matrix was a breeze once I committed to trying the queue, and the same kind of solution worked well for the “clone a graph” problem. I can also see how I could have used it for flood fill.
• I knocked out the k points closest to the origin problem well under the time limit with a very fast solution; it was pure “how do I fill and then sort a custom data structure?”.
• sort.Ints, sort.Slice, and sort.Search are now my very good friends and help immensely.

The bad:

• I am pretty bad at numerical algorithms, like 2sum/3sum, or the inserted interval. I am not intuiting the algorithms at all well.
• I am still not properly catching myself falling down rabbit holes or not being on the right track. I need some heuristics. Maybe “if the if statement are more than 2 deep, you may be off course”.
• If I get a dynamic programming problem, I’m sunk. I might be able to code something if it ends up having a Fibonacci relation, but anything more complex is currently beyond me. The DP solution to the length of the longest substring without repeats is just completely black magic. It works, but I cannot see how it possibly could.

leetcode feels like programming’s highs and lows all condensed into a single highly compressed experience: sometimes it’s absolutely crystal clear, and you just need to get it all typed in (all of the copying of the graph into an adjacency array and recloning the nodes worked the first time; the only thing that didn’t was the link builder, which I had left one line out of).

So sometimes I’m enjoying it, and sometimes I am very much not. I sorely miss a real debugger, but I’m getting along with fmt.Println. I would give a lot to be able to build things the way I prefer, bottom up with tests, but leetcode not gonna work that way.