Ticket #329: optimizations.txt

Fri Jul  3 20:43:01 PDT 2009  kevan@isnotajoke.com
  * Use CachingDict instead of dict in dirnode.py

New patches:

[Use CachingDict instead of dict in dirnode.py
kevan@isnotajoke.com**20090704034301
 Ignore-this: 53f12260176a5170b3599eda54f38e98
] {
hunk ./src/allmydata/dirnode.py 156
         # first we create a MutableFileNode with empty_contents, then use its
         # URI to create our own.
         self._node = self.filenode_class(self._client)
-        empty_contents = self._pack_contents({})
+        empty_contents = self._pack_contents(CachingDict())
         d = self._node.create(empty_contents, keypair_generator, keysize=keysize)
         d.addCallback(self._filenode_created)
         return d
hunk ./src/allmydata/dirnode.py 210
         assert isinstance(data, str)
         # an empty directory is serialized as an empty string
         if data == "":
-            return {}
+            return CachingDict()
         writeable = not self.is_readonly()
hunk ./src/allmydata/dirnode.py 212
-        children = {}
+        children = CachingDict()
         while len(data) > 0:
             entry, data = split_netstring(data, 1, True)
             name, rocap, rwcapdata, metadata_s = split_netstring(entry, 4)
hunk ./src/allmydata/dirnode.py 227
             child = self._create_node(rwcap, rocap)
             metadata = simplejson.loads(metadata_s)
             assert isinstance(metadata, dict)
-            children[name] = (child, metadata)
+            children.set_both_items(name, entry, (child, metadata))
         return children
 
     def _pack_contents(self, children):
hunk ./src/allmydata/dirnode.py 232
         # expects children in the same format as _unpack_contents
-        assert isinstance(children, dict)
+        assert isinstance(children, CachingDict)
         entries = []
         for name in sorted(children.keys()):
hunk ./src/allmydata/dirnode.py 235
-            child, metadata = children[name]
-            assert isinstance(name, unicode)
-            assert IFilesystemNode.providedBy(child), (name,child)
-            assert isinstance(metadata, dict)
-            rwcap = child.get_uri() # might be RO if the child is not writeable
-            if rwcap is None:
-                rwcap = ""
-            assert isinstance(rwcap, str), rwcap
-            rocap = child.get_readonly_uri()
-            if rocap is None:
-                rocap = ""
-            assert isinstance(rocap, str), rocap
-            entry = "".join([netstring(name.encode("utf-8")),
+            entry, metadata = children.get_both_items(name)
+            if entry == None:
+                child, metadata = metadata
+                assert isinstance(name, unicode)
+                assert IFilesystemNode.providedBy(child), (name,child)
+                assert isinstance(metadata, dict)
+                rwcap = child.get_uri() # might be RO if the child is not writeable
+                if rwcap is None:
+                    rwcap = ""
+                assert isinstance(rwcap, str), rwcap
+                rocap = child.get_readonly_uri()
+                if rocap is None:
+                    rocap = ""
+                assert isinstance(rocap, str), rocap
+                entry = "".join([netstring(name.encode("utf-8")),
                              netstring(rocap),
                              netstring(self._encrypt_rwcap(rwcap)),
                              netstring(simplejson.dumps(metadata))])
}

Context:

[setup: the mac-exe build (using py2app) requires macholib>=1.2.
zooko@zooko.com**20090703175702
 Ignore-this: ea489a5076f2c9f20f6bfd2807cc5680
 Here is someone else's description of this problem:
 http://rtmpy.org/ticket/7
] 
[clean up debian packaging: we have control files for etch/lenny/sid, and
"Brian Warner <warner@lothar.com>"**20090703072804
 everything else uses one of those. Add dependency on python-pysqlite2 for
 platforms that use py2.4 by default. Update foolscap dependency to 0.4.1.
] 
[update NEWS to cover all recent changes, sort by end-user importance
Brian Warner <warner@lothar.com>**20090703014303
 Ignore-this: 6ddac78075d7547a19712d505818949c
] 
[Tolerate unknown URI types in directory structures. Part of #683.
Brian Warner <warner@lothar.com>**20090703010749
 Ignore-this: afd0e15e2e39d3b87743ec7ccd87054d
 
 The idea is that future versions of Tahoe will add new URI types that this
 version won't recognize, but might store them in directories that we *can*
 read. We should handle these "objects from the future" as best we can.
 Previous releases of Tahoe would just explode. With this change, we'll
 continue to be able to work with everything else in the directory.
 
 The code change is to wrap anything we don't recognize as an UnknownNode
 instance (as opposed to a FileNode or DirectoryNode). Then webapi knows how
 to render these (mostly by leaving fields blank), deep-check knows to skip
 over them, deep-stats counts them in "count-unknown". You can rename and
 delete these things, but you can't add new ones (because we wouldn't know how
 to generate a readcap to put into the dirnode's rocap slot, and because this
 lets us catch typos better).
] 
[test_client.py: minor refactoring of BASECONFIG usage
Brian Warner <warner@lothar.com>**20090703005450
 Ignore-this: bf19b9ae7edeab293068cce2368f9364
] 
[create_node_from_uri: take both writecap+readcap, move logic out of dirnode.py
Brian Warner <warner@lothar.com>**20090702222537
 Ignore-this: 93051498076e90d3f1dc85161ce8247a
] 
[dirnode.py: prepare to preserve both rwcap+rocap when copying
Brian Warner <warner@lothar.com>**20090702211254
 Ignore-this: f128c02da32f86d7e39527a35dfc2e02
 
 This will make it easier to tolerate unknown nodes safely.
] 
[interfaces.py: wrap some lines to 80cols
Brian Warner <warner@lothar.com>**20090702015728
 Ignore-this: e2c777c1e89a684b43ceabeb0042456c
] 
[cli: webopen: when called with no arguments, open the Welcome page
Brian Warner <warner@lothar.com>**20090701200548
 Ignore-this: ae7d6cb42165d0c751926065378343dd
] 
[mutable repairer: skip repair of readcaps instead of throwing an exception.
Brian Warner <warner@lothar.com>**20090701011343
 Ignore-this: 2c24493426cdc1db8f0e3815ee2c5f87
 This should improve the behavior of #625 a bit: at least all the files will
 get repaired.
] 
[interfaces.py: improve ICheckAndRepairResults docs a bit
Brian Warner <warner@lothar.com>**20090701001925
 Ignore-this: 3a57acfee3487c1e071e135743e42c2a
] 
[repairer.py: wrap to 80cols. No code changes.
Brian Warner <warner@lothar.com>**20090701000047
 Ignore-this: 4a84ac95a849be0656d362882876082a
] 
[setup: require pycryptopp>=0.5.14 if on Windows and with Python>=2.6
zooko@zooko.com**20090630184807
 Ignore-this: f7e9beeb5d5613a7c0ffed14d1dda3c6
] 
[edit NEWS
Brian Warner <warner@lothar.com>**20090630174115
 Ignore-this: c4461a2304fcd45bee95e11418693a18
] 
[NEWS: list all user-visible changes since 1.4.1 . Needs lots of editing.
Brian Warner <warner@lothar.com>**20090630170734
 Ignore-this: f606a5d678d0db8065b9f84e796d59b0
] 
[* find-trailing-spaces.py: exit rc=1 if whitespace found, to be a pre-commit hook
Brian Warner <warner@lothar.com>**20090629224658
 Ignore-this: 1465becc923ee04d170e5ee791cf605e
] 
[use 522-bit RSA keys in all unit tests (except one)
Brian Warner <warner@lothar.com>**20090629223124
 Ignore-this: 7a4c3685683ff9da5ceb2d8cb7b19b7
 
 This reduces the total test time on my laptop from 400s to 283s.
 * src/allmydata/test/test_system.py (SystemTest.test_mutable._test_debug):
   Remove assertion about container_size/data_size, this changes with keysize
   and was too variable anyways.
 * src/allmydata/mutable/filenode.py (MutableFileNode.create): add keysize=
 * src/allmydata/dirnode.py (NewDirectoryNode.create): same
 * src/allmydata/client.py (Client.DEFAULT_MUTABLE_KEYSIZE): add default,
   this overrides the one in MutableFileNode
] 
[remove trailing whitespace
Brian Warner <warner@lothar.com>**20090629200358
 Ignore-this: 7a3756618dcfca0a40acb4c3d15f6440
] 
[remove trailing whitespace from some util classes
Brian Warner <warner@lothar.com>**20090629200341
 Ignore-this: ded84c38885d6aabf2691957dde82bf3
] 
[repairer: raise a better exception when faced with a readonly filenode. Still
Brian Warner <warner@lothar.com>**20090626063230
 Ignore-this: a100005b973a6a57566b943073352828
 produces an error, though.
] 
[mutable/filenode.py: set _writekey to None, rather than leaving it missing
Brian Warner <warner@lothar.com>**20090626062022
 Ignore-this: be111c37dabd6c7aa47abd7bf160926e
 
 This will at least turn the really really weird error when a repair of a
 readonly mutable file is attempted into a merely really weird assertion that
 mentions "repair currently requires a writecap".
] 
[one last tweak
Brian Warner <warner@lothar.com>**20090626052627
 Ignore-this: 8d9d045a1ebe774aeed6f7e1cdd36973
] 
[more 'tahoe cp' docs tweaks
Brian Warner <warner@lothar.com>**20090626051652
 Ignore-this: f594c395007af64da3d7302943ef8855
] 
[cli.py: minor tweaks to test posthooks
"Brian Warner <warner@lothar.com>"**20090626024557] 
[cli.py: update comments on 'tahoe cp' --help a bit
Brian Warner <warner@lothar.com>**20090626023837
 Ignore-this: 4090b06896542fbf260d808e8a991cea
] 
[cli: add some --help text to 'tahoe cp'
warner@lothar.com**20090625235751] 
[make it possible to add/renew-lease from the WUI
Brian Warner <warner@lothar.com>**20090626061824
 Ignore-this: 6fb6a0e8ecfc630464ac37a779078e5
 add add/renew-lease checkbox on the "more info" page check/deep-check forms
] 
[hush pyflakes
warner@lothar.com**20090625021809] 
[Split out NoSharesError, stop adding attributes to NotEnoughSharesError, change humanize_failure to include the original exception string, update tests, behave better if humanize_failure fails.
warner@lothar.com**20090625021707] 
[remove introducer/old.py, will create something similar when the RIIntroducer changes
warner@lothar.com**20090624194038] 
[Basically just a trivial platform detection patch for NetBSD.
midnightmagic@example.com**20090617233647] 
[check_grid.py: print stderr when a subcommand fails
warner@lothar.com**20090623022833] 
[big rework of introducer client: change local API, split division of responsibilites better, remove old-code testing, improve error logging
warner@lothar.com**20090623021047] 
[web/welcome.xhtml: remove trailing whitespace
warner@lothar.com**20090623020909] 
[rrefutil: add trap_deafref utility, to make the callRemote-plus-ignore-DeadReferenceError-plug-log-other-errors pattern easier
warner@lothar.com**20090623020826] 
[PollMixin: snoop trial's error observer to halt the test early if an error is seen. This turns a lot of timeouts into fast failures.
warner@lothar.com**20090623020731] 
[clean up storage_broker interface: should fix #732
warner@lothar.com**20090621235119
 Ignore-this: fb93cd670e809eed2bc123142dd8d4ff
] 
[hush pyflakes with recent FileTooLarge removal
warner@lothar.com**20090621231757
 Ignore-this: 4231b38c7e9091b0577b07ec99ac2df0
] 
[add docs/proposed/GridID.txt (cleaning out some of my old branches)
warner@lothar.com**20090621191204] 
[docs: remove warning about inability to build modules on py2.6 on Windows with mingw, differentiate between clients and servers, reflow to a consistent column width (79), add hint about firewall/NAT docs.
zooko@zooko.com**20090621175005
 Ignore-this: 85e7c1ccb258317ca4dd37917afb48f5
] 
[setup: copy in misc/show-tools-version.py from zfec -- it prints out platform and setuptools versions
zooko@zooko.com**20090621055846
 Ignore-this: 4e144886ab02414bbaaf0295ce2b337
] 
[docs: start updating the NEWS and relnotes.txt files, add Kevan to CREDITS
zooko@zooko.com**20090621055114
 Ignore-this: 35e05a5739549ffa693d55df51ffcfd
] 
[change max filesize limit tests
kevan@isnotajoke.com**20090620212822
 Ignore-this: 38e7c62a308c3c93e79df4bf72f4f675
 
 Instead of testing to see that the previous SDMF filesize limit was being
 obeyed, we now test to make sure that we can insert files larger than that
 limit.
] 
[remove upper limit on SDMF filesize
kevan@isnotajoke.com**20090620213130
 Ignore-this: 5bc48c7421c73827909a17e651799d0c
] 
[immutable/download: instrument do-you-have-block responses to investigate #732
warner@lothar.com**20090621041209] 
[Makefile: add jaunty support, rearrange debian sections in order of release
warner@lothar.com**20090618050502] 
[test_util: add known-answer tests for hashutil tags
warner@lothar.com**20090618045709] 
[tests: bump up a timeout that expired on Zandr's box
zooko@zooko.com**20090613195703
 Ignore-this: 853000f43665396b3734a99f72bd472d
] 
[util: hooray!  A clean implementation of this simple utility!  Black Dew pointed out that the inverse of time.gmtime() is hidden in the "calendar" module.
zooko@zooko.com**20090613160112
 Ignore-this: 8359cc48ca1b8e2793e8b2afe2050cf4
] 
[util: Brian's horrible hack to figure out how much localtime and utctime differ.  Now we'll see if it works on Windows.
zooko@zooko.com**20090612204556
 Ignore-this: 8c36431da4707da76472956c7750ecbd
] 
[util: oops, time.tzset() doesn't work on Windows -- hopefully the new "London" unit test passes on Windows when we skip tzset() on platforms that don't have it
zooko@zooko.com**20090612000920
 Ignore-this: 7d314b8334cfa3f65f9635e3d3eb727e
] 
[setup: edit install.html to warn Windows users away from Python v2.6
zooko@zooko.com**20090611225506
 Ignore-this: 89ad63eab49ede883ef92f2de5b5fc54
] 
[util: fix time_format.iso_utc_time_to_seconds() so that it works even in London
zooko@zooko.com**20090611221129
 Ignore-this: 14dbb9840587797f848226fdb3645c08
] 
[setup: run the same "make quicktest" on Windows as on non-Windows
zooko@zooko.com**20090611193214
 Ignore-this: d0c78377d3892373ec3d91e9e98bd8bd
 I checked and it behaves about as well on Windows as the previous version did.
] 
[tests: significantly increase timeouts that triggered on Zandr's ARM box
zooko@zooko.com**20090610161043
 Ignore-this: 2b3c556c5166a8267b4b15664d3aadfb
] 
[test: multiple by 10 or so all timeouts that Zandr's ARM box just overran
zooko@zooko.com**20090610125639
 Ignore-this: bf62f063ab46814fd78de55a5fbc9d84
] 
[tests: bump up timeout on a test that timed out on draco
zooko@zooko.com**20090610044628
 Ignore-this: f598b98cbae44dc947937c6ca54c10cb
] 
[tests: raise the timeout for test_cli since Zandr's ARM machine totally burst through the old one
zooko@zooko.com**20090609210509] 
[test_cli.Backup: increase timeout massively, it takes 1200s on zandr's ARM linkstation
warner@lothar.com**20090609052801] 
[tests: double the timeouts on some tests which time-out on Francois's box
zooko@zooko.com**20090609021753
 Ignore-this: b2727b04402f24a9b9123d2f84068106
] 
[tests: bump up timeouts so that the tests can finish before timeout on Francois's little arm box
zooko@zooko.com**20090608225557
 Ignore-this: fb83698338b2f12546cd3e1dcb896d34
] 
[tests: increase timeouts on some other tests that timed-out on Francois's arm box
zooko@zooko.com**20090605143437
 Ignore-this: 2903cc20d914fc074c8d7a6c47740ba6
] 
[tests: bump up the timeout on a bunch of tests that took longer than the default timeout (120s) on François Lenny-armv5tel
zooko@zooko.com**20090605031444
 Ignore-this: 84d67849b1f8edc88bf7001e31b5f7f3
] 
[backup: remove the --no-backupdb command, the handling of "can't import sqlite", and the related tests, and change an error message to more correctly indicate failure to load the database from disk rather than failure to import sqlite module
zooko@zooko.com**20090604173131
 Ignore-this: 8200a9fdfc49243c280ecd1d0c44fa19
 Fixes #728.
] 
[setup: require pysqlite >= v2.0.5. if we are running on Python < 2.5
zooko@zooko.com**20090604154548
 Ignore-this: cf04f46079821df209d01dad2e24b40b
] 
[setup: add pysqlite and sqlite to get_package_versions()
zooko@zooko.com**20090604153728
 Ignore-this: a1dea7fabeab2b08fb0d8d462facdb4d
] 
[more refactoring: move get_all_serverids() and get_nickname_for_serverid() from Client to storage_broker
warner@lothar.com**20090602030750] 
[more storage_broker refactoring: downloader gets a broker instead of a client,
warner@lothar.com**20090602022511
 use Client.get_storage_broker() accessor instead of direct attribute access.
] 
[test_runner.py: remove test_client_no_noise: the issue in question is
warner@lothar.com**20090601225007
 ticketed in http://divmod.org/trac/ticket/2830 and doesn't need a Tahoe-side
 change, plus this test fails on win32 for unrelated reasons (and test_client
 is the place to think about the win32 issue).
] 
[remove plaintext-hashing code from the helper interface, to close #722
warner@lothar.com**20090601224916
 and deny the Helper the ability to mount a partial-information-guessing
 attack. This will probably break compatibility between new clients and very
 old (pre-1.0) helpers.
] 
[start to factor server-connection-management into a distinct 'StorageServerFarmBroker' object, separate from the client and the introducer. This is the starting point for #467: static server selection
warner@lothar.com**20090601210604] 
[mutable: catch and display first error, so code bugs which break all servers get displayed better
warner@lothar.com**20090601210407] 
[misc/run-with-pythonpath.py: exec() the child (on unix), to remove the intermediate process
warner@lothar.com**20090601210137] 
[docs: small edit to about.html
zooko@zooko.com**20090528233422
 Ignore-this: 1cfbb1f8426ed6d63b2d3952e4464ddc
] 
[docs: add links to Tahoe-LAFS for Paranoids and Tahoe-LAFS for Corporates in about.html
zooko@zooko.com**20090528232717
 Ignore-this: 7b70baa700d6b6f6e9ceec4132efe5
] 
[docs: edit about.html and include network-and-reliance-topology.png (loaded from http://allmydata.org )
zooko@zooko.com**20090527150916
 Ignore-this: 44adc61cde8ced8be2f0a7dfc7d95dad
] 
[docs: a few more edits to network-and-reliance-topology.svg
zooko@zooko.com**20090527150458
 Ignore-this: 2eac8c33fe71be25ff809b399c6193c1
] 
[docs: update network-and-reliance-topology.svg for beauty and clarity
zooko@zooko.com**20090527031123
 Ignore-this: 5510914849771900ac29b4312470d84
] 
[docs: update NEWS, relnotes.txt, CREDITS to mention WUI Style
zooko@zooko.com**20090526233654
 Ignore-this: 72d16ec833bc4a22af23d29ea1d5ff8b
] 
[Modify markup of Tahoe web pages to be more amenable to styling; some minor changes of wording.
Kevin Reid <kpreid@mac.com>**20090526232545
 Ignore-this: 8845937f0df6c7ddc07abe3211428a6f
] 
[Tweak wording in directory page: not-read-only is "modifiable", mention creating a directory _in this directory_.
Kevin Reid <kpreid@mac.com>**20090526232414
 Ignore-this: f006ec52ba2051802e025a60bcface56
] 
[Comment on duplication of code/markup found during styling project.
Kevin Reid <kpreid@mac.com>**20090503203442
 Ignore-this: a4b7f9f0ab57d2c03be9ba761be8d854
] 
[Add CSS styles to spiff up the Tahoe WUI's appearance, particularly the welcome page and directories.
Kevin Reid <kpreid@mac.com>**20090503203142
 Ignore-this: 5c50af241c1a958b5180ef2b6a49f626
] 
[Link all Tahoe web pages to the /tahoe_css stylesheet which already exists.
Kevin Reid <kpreid@mac.com>**20090503202533
 Ignore-this: 2ea8d14d3168b9502cf39d5ea3f2f2a8
] 
[Fix broken link from Provisioning to Reliability page.
Kevin Reid <kpreid@mac.com>**20090501191050
 Ignore-this: 56dc1a5e659b70cc02dc4df7b5d518cd
] 
[docs: network-and-reliance-topology.svg: nicer server icons, mv out of the "specifications" subdir
zooko@zooko.com**20090526165842
 Ignore-this: 8f47ab3a0ab782c1f0d46e10bcaebe5b
] 
[docs: update network-and-reliance-topology.svg
zooko@zooko.com**20090526163105
 Ignore-this: 2b864b4ed8743d4a15dfbb7eff3fa561
] 
[accounting-overview.txt: more edits
warner@lothar.com**20090523190359] 
[accounting-overview.txt: small edits
warner@lothar.com**20090523184011] 
[_auto_deps.py: require foolscap-0.4.1, which adds an important fix for py2.4
warner@lothar.com**20090523011103] 
[immutable/encode.py: tolerate immediate _remove_shareholder by copying the
warner@lothar.com**20090522184424
 landlord list before iterating over it. This can probably only happen in unit
 tests, but cleaning it up makes certain test failures easier to analyze.
] 
[switch to using RemoteException instead of 'wrapped' RemoteReferences. Should fix #653, the rref-EQ problem
warner@lothar.com**20090522004632] 
[switch all foolscap imports to use foolscap.api or foolscap.logging
warner@lothar.com**20090522003823] 
[_auto_deps.py: bump our foolscap dependency to 0.4.0, since I'm about to start using its new features
warner@lothar.com**20090522002100] 
[test_runner.py: fix minor typo
warner@lothar.com**20090520033620] 
[setup: fix bug (wrong import) in error message, as noticed by pyflakes
zooko@zooko.com**20090519195642
 Ignore-this: f1b9f8c00b46c1b5f2f20e5fc424f341
] 
[setup: fix trivial bug in recent patch to test base64.py at startup
zooko@zooko.com**20090519195129
 Ignore-this: f6be038f74b53ca69e7109fe34adfbc
] 
[setup: make Tahoe exit at startup with a useful error message if the base64.py module is buggy (fixes part of #710)
zooko@zooko.com**20090519194555
 Ignore-this: aa4d398235ddca8d417d61c9688e154
] 
[test_introducer.py: add a test for the python2.4.0/2.4.1 bug in base64.b32decode
warner@lothar.com**20090519034101] 
[immutable WriteBucketProxy: use pipeline to speed up uploads by overlapping roundtrips, for #392
warner@lothar.com**20090518234422] 
[util/pipeline.py: new utility class to manage size-limited work pipelines, for #392
warner@lothar.com**20090518234326] 
[docs: add a diagram that I'm about to show to the Boulder Linux Users Group: network-and-reliance-topology.svg
zooko@zooko.com**20090514232059
 Ignore-this: 2420c0a7c254c9f0f2349d9130490d33
] 
[tests: mark test_runner as coded in utf-8 instead of ascii
zooko@zooko.com**20090507223151
 Ignore-this: ccf1ba9e5a9b53602701a36f9fdb545e
] 
[tests: raise timeout on test_runner.RunNode.test_introducer from 120s to 240s, since it hit the 120s time-out on François Lenny-armv5tel
zooko@zooko.com**20090507215012
 Ignore-this: ba18fe6832ba255d4971e8f623ed7da5
] 
[setup: fix comment in setup.py
zooko@zooko.com**20090507215003
 Ignore-this: c46ef664630d52733138ef7fbc551c1c
] 
[docs: how_to_make_a_tahoe_release.txt: a couple of small edits
zooko@zooko.com**20090507214932
 Ignore-this: ae92aa835ad369f4b9e6e49d681957a3
] 
[.darcs-boringfile: also ignore .gitignore
warner@allmydata.com**20090415210550
 Ignore-this: d29db314a1e506f6240859559436b4c3
] 
[.darcs-boringfile: ignore .git, I'm starting to play around with it
warner@allmydata.com**20090415205929
 Ignore-this: 89234453516483c9586cd6e1351e88b5
] 
[fix quicktest: stop using setuptools, add misc/run-with-pythonpath.py, to make it run faster
warner@lothar.com**20090414201400] 
[TAG allmydata-tahoe-1.4.1
zooko@zooko.com**20090414025636
 Ignore-this: de78fc32364c83e9f4e26b5abcfdea4a
] 
Patch bundle hash:
4205616ff5ad322114503f4366e7f7ff9c780916