How was this intergral derived from Bayes' Rule in David Heckerman's Bayesian Network paper? I am trying to follow this paper titled "A Tutorial on Learning With Bayesian Networks" by Microsoft researcher David Heckerman. 
In it I am unable to figure out how he got to Equation 2 from Equation 1. There is an assumption being made that I can't figure out.
$$p(\theta|D,\xi) = \frac{p(\theta|\xi ) p(D|\theta,\xi)}{p(D|\xi)} \tag{Equation 1}$$
$$p(D|\xi) = \int p(D|\theta,\xi ) \, p(\theta|\xi) \, \mathrm{d}\theta  \tag{Equation 2}$$
In my attempt to try to help myself, I have found myself learning about contour integration and normalized Gaussian distributions.  But I still don't see how the author got to equation 2. Or why he would include that in the paper, as it was never mentioned again.
References
Heckerman, David (1995). "A tutorial on learning with Bayesian networks." Microsoft Research Technical Report MSR-TR-95-06.
 A: It comes from the following:
$$p(D \vert \xi) = \int_\Theta p(D, \theta | \xi)d\theta = \int_\Theta p(D \vert \theta, \xi) p(\theta \vert \xi) d\theta.$$
This isn't actually coming from Eqn. 1 (which is just Bayes rule). It is just following the basic rules of probability.
To be more explicit: there are two steps happening here.
The first is that we can integrate $p(D, \theta \vert \xi)$ over $\theta$ to get $p(D \vert \xi)$. We are going from a joint conditional distribution to a marginal conditional distribution.
The second step is using the laws of conditional probability to write $p(D, \theta \vert \xi) = p(D \vert \theta, \xi) p(\theta \vert \xi)$. We then integrate this.
$$ \ $$
Update: I just actually looked at the paper that you linked to, so here's an expanded explanation.
Equation 1 is (conditional) Bayes rule, which relates the posterior $p(\theta \vert D, \xi)$ to the prior $p(\theta \vert \xi)$, the likelihood $p(D \vert \theta, \xi)$, and the evidence $p(D \vert \xi)$. Bayesian inference flows from this.
$p(D \vert \xi)$ is a normalizing constant (with respect to $\theta$, the thing we care about), and generally is not of interest in and of itself, although we need it if we want to evaluate our posterior at certain points (note that one of the appealing things about the Metropolis-Hastings algorithm for MCMC is that you don't need to compute the evidence). The reason it's nice to write $p(D \vert \xi)$ the way that he did is because then we have expressed the posterior entirely in terms of the likelihood and the prior (which are things we very much do care about). So once we have a prior and a likelihood we can (in principle) compute the posterior.
