Bug 1884327 - faster perfect hash r=ahochheiden

While we're at it, also remove the `six` layer, which slightly impacts
some callers.

Depends on D204009

Differential Revision: https://phabricator.services.mozilla.com/D204010

xpcom/components/gen_static_components.py

@@ -920,14 +920,18 @@ def gen_substs(manifests):
 
     cid_phf = PerfectHash(modules, PHF_SIZE, key=lambda module: module.cid.bytes)
 
-    contract_phf = PerfectHash(contracts, PHF_SIZE, key=lambda entry: entry.contract)
+    contract_phf = PerfectHash(
+        contracts, PHF_SIZE, key=lambda entry: entry.contract.encode()
+    )
 
     js_services_phf = PerfectHash(
-        list(js_services.values()), PHF_SIZE, key=lambda entry: entry.js_name
+        list(js_services.values()), PHF_SIZE, key=lambda entry: entry.js_name.encode()
     )
 
     protocol_handlers_phf = PerfectHash(
-        list(protocol_handlers.values()), TINY_PHF_SIZE, key=lambda entry: entry.scheme
+        list(protocol_handlers.values()),
+        TINY_PHF_SIZE,
+        key=lambda entry: entry.scheme.encode(),
     )
 
     js_services_json = {}
@@ -945,7 +949,7 @@ def gen_substs(manifests):
     substs["contract_count"] = len(contracts)
     substs["protocol_handler_count"] = len(protocol_handlers)
 
-    substs["default_protocol_handler_idx"] = protocol_handlers_phf.get_index("default")
+    substs["default_protocol_handler_idx"] = protocol_handlers_phf.get_index(b"default")
 
     gen_module_funcs(substs, module_funcs)
 

xpcom/ds/tools/perfecthash.py

@@ -24,18 +24,6 @@
 import textwrap
 from collections import namedtuple
 
-import six
-from mozbuild.util import ensure_bytes
-
-
-# Iteration over bytestrings works differently in Python 2 and 3; this function
-# captures the two possibilities. Returns an 'int' given the output of iterating
-# through a bytestring regardless of the input.
-def _ord(c):
-    if six.PY3:
-        return c
-    return ord(c)
-
 
 class PerfectHash(object):
     """PerfectHash objects represent a computed perfect hash function, which
@@ -88,7 +76,7 @@ def __init__(self, entries, size, validate=True, key=lambda e: e[0]):
 
         for bucket in buckets:
             # Once we've reached an empty bucket, we're done.
-            if len(bucket.entries) == 0:
+            if not bucket.entries:
                 break
 
             # Try values for the basis until we find one with no conflicts.
@@ -103,7 +91,7 @@ def __init__(self, entries, size, validate=True, key=lambda e: e[0]):
                     # There was a conflict, try the next basis.
                     basis += 1
                     idx = 0
-                    del slots[:]
+                    slots.clear()
                     assert basis < self.U32_MAX, "table too small"
                 else:
                     slots.append(slot)
@@ -127,15 +115,16 @@ def _hash(cls, key, basis=FNV_OFFSET_BASIS):
         32-bit FNV is used for indexing into the first table, and the value
         stored in that table is used as the offset basis for indexing into the
         values table."""
-        for byte in memoryview(ensure_bytes(key)):
-            obyte = _ord(byte)
+        FNV_PRIME = cls.FNV_PRIME
+        U32_MAX = cls.U32_MAX
+        for obyte in memoryview(key):
             basis ^= obyte  # xor-in the byte
-            basis *= cls.FNV_PRIME  # Multiply by the FNV prime
-            basis &= cls.U32_MAX  # clamp to 32-bits
+            basis *= FNV_PRIME  # Multiply by the FNV prime
+            basis &= U32_MAX  # clamp to 32-bits
         return basis
 
     def key(self, entry):
-        return memoryview(ensure_bytes(self._key(entry)))
+        return memoryview(self._key(entry))
 
     def get_raw_index(self, key):
         """Determine the index in self.entries without validating"""
@@ -145,7 +134,7 @@ def get_raw_index(self, key):
     def get_index(self, key):
         """Given a key, determine the index in self.entries"""
         idx = self.get_raw_index(key)
-        if memoryview(ensure_bytes(key)) != self.key(self.entries[idx]):
+        if memoryview(key) != self.key(self.entries[idx]):
             return None
         return idx
 
@@ -334,7 +323,7 @@ def gen_jslinearstr_getter(
                             not in the table."""
 
         assert all(
-            _ord(b) <= 0x7F for e in self.phf.entries for b in self.phf.key(e)
+            b <= 0x7F for e in self.phf.entries for b in self.phf.key(e)
         ), "non-ASCII key"
 
         if return_type is None: