It just adds extra complexity and is in conflict for our plans for
Lean4. Moreover, in our experiments it impacts negatively on
performance: master and lean4 branches. The negative impact has been
confirmed by @kha too.
This is the correct fix for the id declaration pretty printing
discrepancy reported by Daniel.
TODO: decide whether we need another eq-mode where names are ignored.
For example, in blast, it makes sense to increase sharing by ignoring
binder names.
We simulate it in the following way:
1- An opaque 'let'-expressions (let x : t := v in b) is encoded as
((fun (x : t), b) v)
We also use a macro (let-macro) to mark this pattern.
Thus, the pretty-printer knows how to display it correctly.
2- Transparent 'let'-expressions are eagerly expanded by the parser.
Signed-off-by: Leonardo de Moura <leonardo@microsoft.com>
We do not enforce full hash-consing because we would need to synchronize
the access to the hashtable/cache.
Signed-off-by: Leonardo de Moura <leonardo@microsoft.com>
