the CFL pumping constant
Every pumping lemma comes with a threshold: a length p such that ONCE a string in the language is at least that long, the pumpable structure is guaranteed to appear. For context-free languages this threshold is the pumping constant. Below it, short strings may have no repeatable piece; at or above it, repetition is forced — and that forced repetition is what the lemma exploits.
Where does p come from? Put the language's grammar in Chomsky normal form, where every rule produces either two variables or one terminal, so parse trees are BINARY. If a grammar has b variables (nonterminals), a parse tree whose longest root-to-leaf path has length at most b can yield a string of length at most 2^(b-1). So choose p = 2^b (or any value larger than the longest yield of a path-bounded tree). Then any string of length >= p must have a parse tree with a path longer than b nodes, and on a path with more than b labelled nodes some VARIABLE must repeat — that is the pigeonhole step. The repeated variable is exactly what gives the u v x y z' split, with v x y kept short (bounded by p) because it is the yield of a bounded subtree near the bottom of that path.
In practice you almost never compute p. The lemma says only that SOME p exists; in a proof you treat it as an unknown fixed number, choose a string whose length depends on p (so it is guaranteed long enough), and derive a contradiction without ever pinning p down. The constant is conceptual scaffolding — its existence, traced back to the finite number of variables in a CNF grammar, is the whole point; its exact numeric value rarely matters.
If a CNF grammar has b = 4 variables, take p = 2^4 = 16. Any string in its language of length 16 or more must, in any parse tree, have a root-to-leaf path repeating one of the 4 variables — by pigeonhole, since the path has more than 4 labelled internal nodes. That repeat yields the pumpable u v x y z' decomposition.
p ties to the height of a Chomsky-normal-form parse tree: enough variables on a path force one to repeat.
p depends on the grammar, not just the language, and you rarely need its exact value — proofs treat it as a fixed-but-unknown constant. Its existence comes from a CNF grammar having only finitely many variables.