Solve a Linear Equation

JavaScript ES6, 246 bytes

Still some golfing to be done, but at least it's a solution!

C=a=>new Function("x","return "+a.replace(/(\d)x/g,"$1*x"));n=>{n=n.split("=");t=Math.abs,r=C(n[0]),c=C(n[1]),a=0,i=r(a)-c(a);a++;v=r(a)-c(a);o=t(i)<t(v)?-1:1;for(u=1/0;r(a)!==c(a);)a+=o,e=t(r(a)-c(a)),e>u&&(u=1/0,o/=10),u=Math.min(e,u);return a}

Name the function n=>{n=n.split("=")... to use it.

Hyper-ungolfed:

function solveLinear(equation){
    equation = equation.split("=");
    var abs = Math.abs;
    var LHS = convertToFunction(equation[0]), RHS = convertToFunction(equation[1]);
    var pivot = 0;
    var dir;
    var dir1 = LHS(pivot) - RHS(pivot);
    pivot++;
    var dir2 = LHS(pivot) - RHS(pivot);
    if(abs(dir1)<abs(dir2)) dir = -1;
    else dir = 1;
    var dif, minDif = Infinity;
    while(LHS(pivot) !== RHS(pivot)){
        pivot += dir;
        dif = abs(LHS(pivot) - RHS(pivot));
        if(dif > minDif){
            minDif = Infinity;
            dir /= 10;
        }
        minDif = Math.min(dif, minDif);
        console.log(pivot,dir,dif,minDif);
    }
    return {
        x: pivot,
        LHS: LHS,
        RHS: RHS
    };
}

This uses a pivot approach. (I'm not sure if this is what the algorithm is called, just a name I invented.) It first gathers which direction to search for from zero (i.e., which way the slopes of the two sides of the equations will intersect) and looks for the value. Once it finds a point of minimal difference, it goes to that point and decreases the search increment. This eventually yields as precise of a solution we need.

JavaScript (Node.js), 106 93 bytes

a=>eval(`f=x=>${a[R='replace'](/(\d)x/g,"$1*x")[R]("=","-(")[R](/-/g,"+-")})`)(0)/(f(0)-f(1))

Try it online!

-13 bytes thanks to @tsh

Ungolfed:

var h=a=>{
  a=a.replace(/(\d)x/g,"$1*x").replace("=","-(").replace("--","- -"); //get into an eval-able form
  var f=x=>eval(a+")");
  var df=(f(1)-f(0))/(1-0) //derivative or slope of the function
  var x=0;
  return x-(f(x)/df); //newton's method
}

Explaination:

This solution works by Newton's method for finding roots. The code subtracts the right hand side of the equation from the left hand side, such that when f(x)=0, x will equal the value we are solving for. Therefore, when we find the root of this new function, it will be our desired x value. Then it finds the derivative f'(x) by finding the slope between two points on the function. Then, the values are simply inserted into Newton's method which states for an approximation of the root x, x=x-(f(x)/f'(x)) (in the code, we use 0 as an initial x value). Since this finds the roots, it finds our x value. And since the equation is guaranteed to be linear, the approximation will be exact.