A 2SAT solver, written in Rust. The 2SAT language is the set of 2CNFs $\phi$ that are satisfiable. Specifically, $\phi\in 2SAT$ iff $\phi$ is a Boolean formula for which there exists at least one assignment of variables that makes $\phi$ true, and $\phi$ is of the form $\phi = C_1\wedge C_2\wedge ... \wedge C_n$ where each $C_i$ is a disjunction of at most 2 literals (i.e., an expression $a$ or $(a\vee b)$ where $a$ and $b$ are of the form $x_i$ or $\neg x_i$ for any input Boolean variable $x_i$).

The generalized problem $SAT$, of satisfiable Boolean formulas that are not restricted to be 2CNFs, is well known to be $NP$-Complete. However, 2SAT can be solved in linear time, and thus is definitely in $P$.

An input formula is specified using ~ for logical negation, & for logical AND, and | for logical OR. Variables may be any alphanumeric string, not starting with a number. Then, the input must be specified as a 2CNF, where clauses (ORs) with 2 variables are enclosed in parentheses, and all clauses are conjoined by &. & must not appear within parentheses, only | can. | cannot appear outside parentheses. Parentheses may not be nested. Some valid formula strings:

"(a|b)&~c&(~a|~d)&(a|a)&c"
"(x1|x2)&(x3|~x4)"

Some invalid formula strings (they are not 2CNFs):

"a|b"
"(a&b)|(c&d)"
"(a|b|c)"
"a&(b|c)&((c|a)&d|e)"

Note that my parser removes all whitespace, so an input like "hello there & (a | hello there)" (which is objectively terrible variable naming), is treated as "hellothere&(a|hellothere)" where the two variables are a and hellothere.

Then, if your correctly specified formula is <phi>, check if it's satisfiable with

$ cargo run <phi>

For example,

$ cargo run "(a|b)&(~a|d)&(b|c)&(~b|d)&(~b|e)&(~d|~c)&(~e|~d)"
Formula is not SAT

$ cargo run "(x1|x2) & (x1|~x3) & (~x1|~x2) & (x1|x4) & (~x1|~x5)"
Formula is SAT with assignment {"x4": true, "x3": false, "x1": false, "x2": true, "x5": false}

If you are curious enough to search through the git history of this repository you will find that I initially implemented this 2SAT solver using the logical rule of resolution. However, it ran in $O(n^4)$ time or something like that, incredibly inefficient (it was still polytime though!). Regardless, after learning about a linear time solution to 2SAT using strongly connected components of its implication graph (see Wikipedia), I decided to convert my solution to this more efficient one that other people had figured out, something more respectable to publish online.