100k LOC would be, let's be generous, about 8 million characters (80/line average, which is high). 8GB of RAM for 8MB of input seems excessive to me. What's going on with that IDE that it needs so much for something so small? Even if that were 8 million distinct tokens/symbols in the input, that's still 1KB/token of memory consumed.
Except that is only what an IDE does at the very minimum. An IDE loads plugins, provides intellisense, documentation, potentially loads documentation of all the libraries, watches files, displays Git status, parses your code for syntax highlighting etc. etc.
You're not opening a TXT file with 100K lines in it.
For every key stroke a new immutable string is created with the content of the file/state. Plus data structure for AST/static analysis of every file in the project including dependencies. Plus another copy of the file for rendering with syntax highlight. One would think all this copying of data is very inefficient, but computers are good at moving data around. The most costly is rendering the glyphs.
