← Back to branch summary

~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/btree_index.py

  • Committer: John Arbash Meinel
  • Date: 2008-08-20 22:12:00 UTC
  • mto: This revision was merged to the branch mainline in revision 3644.
  • Revision ID: john@arbash-meinel.com-20080820221200-1pnk090rjwbydavn
Start working on an alternate way to track compressed_chunk state.
I tried doing a copy() before flush() but that is quite expensive.
Both because it now does copy() but it also does a bigger flush.
Next I'll try tracking 2 objects.

Show diffs side-by-side

added added

removed removed

Lines of Context:
bzrlib/btree_index.py
239
239
            except StopIteration:
240
240
                current_values[pos] = None
241
241
 
242
 
    def _add_key(self, string_key, key_line, rows):
 
242
    def _add_key(self, string_key, line, rows):
243
243
        """Add a key to the current chunk.
244
244
 
245
245
        :param string_key: The key to add.
246
 
        :param key_line: The fully serialised key and value.
 
246
        :param line: The fully serialised key and value.
247
247
        """
248
248
        if rows[-1].writer is None:
249
249
            # opening a new leaf chunk;
250
250
            for pos, internal_row in enumerate(rows[:-1]):
251
251
                # flesh out any internal nodes that are needed to
252
 
                # preserve the high of the tree
 
252
                # preserve the height of the tree
253
253
                if internal_row.writer is None:
254
254
                    length = _PAGE_SIZE
255
255
                    if internal_row.nodes == 0:
256
256
                        length -= _RESERVED_HEADER_BYTES # padded
257
 
                    internal_row.writer = chunk_writer.ChunkWriter(
258
 
                        length, 0)
 
257
                    internal_row.writer = chunk_writer.ChunkWriter(length, 0)
259
258
                    internal_row.writer.write(_INTERNAL_FLAG)
260
259
                    internal_row.writer.write(_INTERNAL_OFFSET +
261
260
                        str(rows[pos + 1].nodes) + "\n")
265
264
                length -= _RESERVED_HEADER_BYTES # padded
266
265
            rows[-1].writer = chunk_writer.ChunkWriter(length)
267
266
            rows[-1].writer.write(_LEAF_FLAG)
268
 
        if rows[-1].writer.write(key_line):
 
267
        if rows[-1].writer.write(line):
269
268
            # this key did not fit in the node:
270
269
            rows[-1].finish_node()
271
 
            next_key_line = string_key + "\n"
 
270
            key_line = string_key + "\n"
272
271
            new_row = True
273
272
            for row in reversed(rows[:-1]):
274
273
                # Mark the start of the next node in the node above. If it
275
274
                # doesn't fit then propogate upwards until we find one that
276
275
                # it does fit into.
277
 
                if row.writer.write(next_key_line):
 
276
                if row.writer.write(key_line):
278
277
                    row.finish_node()
279
278
                else:
280
279
                    # We've found a node that can handle the pointer.
296
295
                new_row.writer.write(_INTERNAL_FLAG)
297
296
                new_row.writer.write(_INTERNAL_OFFSET +
298
297
                    str(rows[1].nodes - 1) + "\n")
299
 
                new_row.writer.write(next_key_line)
300
 
            self._add_key(string_key, key_line, rows)
 
298
                new_row.writer.write(key_line)
 
299
            self._add_key(string_key, line, rows)
301
300
 
302
301
    def _write_nodes(self, node_iterator):
303
302
        """Write node_iterator out as a B+Tree.
326
325
                # First key triggers the first row
327
326
                rows.append(_LeafBuilderRow())
328
327
            key_count += 1
 
328
            # TODO: Flattening the node into a string key and a line should
 
329
            #       probably be put into a pyrex function. We can do a quick
 
330
            #       iter over all the entries to determine the final length,
 
331
            #       and then do a single malloc() rather than lots of
 
332
            #       intermediate mallocs as we build everything up.
 
333
            #       ATM 3 / 13s are spent flattening nodes (10s is compressing)
329
334
            if self.reference_lists:
330
335
                flattened_references = ['\r'.join(['\x00'.join(reference)
331
336
                                                   for reference in ref_list])