anandvarma / namegen Goto Github PK

• Features
• API documentation
• Examples
• CLI screenshots and usage

Every call to GetForId() creates a new strings.Builder object. Although this allocation should happen on the stack and is not a concern as such, see if it would help to Reset() the builder instead of creating a new one each time. Might come in useful particularly when we need not return the string such as with #9

• Github actions to run unit tests after each commit

See if it makes sense to implement the io.Reader interface for zero copy use-cases.

Hello, I've spotted a bug where you cannot change the default delim for the name generator.
I will send over a PR soon

The getPostfixId implementation first encodes the random number into base-10 or base-36 and then returns a substring of it.
These encoded intermediary strings are 24 Bytes long in the common case. But the returned strings are typically 1-8 Bytes, leading to a large amount of fragmentation and trapped memory.

Today there exists logic to detect duplicates within a given dictionary of words. We need to add a test to spot duplicates across dictionaries too. For instance colors might show up in the color and adjective dictionaries which when used together could end up creating names with duplicate words in them.

Since we know the size of all the random strings as well as the postfix before hand, we can preallocate the builder to cut down on resizes and to make the memory allocations more deterministic.

With just 3 entries, it is fairly repetitive at this point. Populate the dictionary!

Add benchmarks and optimize runtime and memory allocations.

GetForId() makes 3 allocations without prefix and 5 with. This is mostly down to initialized array appends and the call to strings.Join. Also evaluate and use strings.Builder to cut down on concatenation related allocations.

Benchmark outputs for reference:
BenchmarkGet-4 4428757 256.4 ns/op 65 B/op 3 allocs/op
BenchmarkGetWithPostfix-4 2684820 440.3 ns/op 162 B/op 5 allocs/op