Kaushik Kulkarni

Results 78 comments of Kaushik Kulkarni

I hit this too. My current (ugly) workaround is launching `sudo ktikz` from the terminal.

@jfmcarreira: `ls -l` in `/tmp` works fine. Some more info: Running Ubuntu 20.04, Kitkz 0.12

For the log: ``` [LaTeX] Error: run failed. Command: pdflatex -shell-escape -halt-on-error -file-line-error -interaction nonstopmode -output-directory /tmp/ktikz-JWownQ /tmp/ktikz-JWownQ/temptikzcode.tex [LaTeX] Warning: could not load LaTeX log file. Log file: /tmp/ktikz-JWownQ/temptikzcode.log ```...

> Pymbolic flattens those, The AST is correct (i.e. unflattened) but I think the `StringifyMapper` is to be blamed here.: ```python >>> a * (b * c) Product((Variable('a'), Product((Variable('b'), Variable('c')))))...

> No, no concrete use case. Well given floating point arithmetic is not associative we can always buggy behavior associated with this in a generated code ;).

> I can think of use cases where I'd want the parentheses and where I wouldn't. (Math vs CS, really.) Would an option be OK? I think its unsafe to...

I think the auto-flattening is just fine. See ```python >> import pymbolic.primitives as p >>> a = p.Variable("a") >>> b = p.Variable("b") >>> c = p.Variable("c") >>> (a * b)...

> Is this worth pursuing given that compile times are suffering because of expression traversal complexity in pymbolic? I think this is useful. IIUC, this PR quite significantly brings down...

However, I wonder how the loopy CI failures could be fixed? By creating a derived version of every pymbolic expression type in loopy so that they can use loopy-specific EqualityMapper?

Oops had missed the branch in the description. I think those will work fine. Thanks!