A search begins in the place most likely to have the positive result. If you dropped your keys while jogging, you would assign a much higher probability to the area on your route rather than the paths you didn’t tread. If you’re uncertain about where you ran, perhaps you passed through a wooded area with no clear trail, then your probability will be diffused, diluted among all possible routes. And because of this, you will probably put off searching in this area last, since you’re the least likely to be successful when searching in this area.
In p-search you will be assigning probabilities to the candidates you’re searching on, though it has to be in a much more constrained way. As mentioned earlier, each candidate has different properties, and these properties are used to determine which types of functions can be used to assign probabilities. In p-search, these are called “priors,” since, prior to actually searching, you are asserting what you believe to be important.
If for example you’re looking for a particular function, you’ll conjure in your mind words which you believe the function will contain. You can then perform a text query in p-search which will assign higher probabilities to files containing these words. If you believe a file was modified at a certain time, you can assert such a prior, and the closer a file was modified to that time, the higher a probability is given to it. As mentioned in Candidate Mappings ??LINK, additional fields can be extracted from candidates like a name or author. These fields can too become the target of a search.