描述
In https://github.com/jlfwong/speedscope/pull/188, I introduced TextDecoder as a way of avoiding performance issues with building up a huge string slowly. There are ways to make the fallback path faster and use less memory as well:
From @evanw:
The problem is that JavaScript JITs such as V8 optimize repeated string appends by building up a linked list in memory and only constructing the full string when the string is first used. So you probably want to periodically flatten the string in your loop to avoid running out of memory.
For future reference: one way to get Chrome to evaluate the string is to do
str[0];. That doesn't appear to do anything but actually has the side effect of evaluating the string.
This fallback path is probably only used in Edge: https://caniuse.com/#feat=textencoder