Showing pickle disassembly can sometimes be useful to analyse details of
the pickle content. We realized that in some data structures used in
ERP5 the same string was saved multiple times in the same pickle and by
using the exact same string (ie. for which `s1 is s2` is True), the
pickle will have the string only once and pickles are a bit smaller. For
more reference, the context was
nexedi/erp5!1560 (comment 154825)

This introduces a new --pretty option that we will be able to extend
later with more output formats.
Co-authored-by: Kirill Smelkov <kirr@nexedi.com>
Reviewed-on: nexedi/zodbtools!22

@kirr wrote (nexedi/zodbtools!19 (comment 129442))

    For the reference - contrary to ZODB5, restore tests on ZODB4 are currently
    [broken](https://nexedijs.erp5.net/#/test_result_module/20210317-B3AC205A/2).
    Restored file is not bit-to-bit identical to the original.

    The problem is that on commit/restore, we need to save
    user/description/extension. For extension `zodbdump.Transaction` provides
    .extension_bytes, which ZODB5 uses to save its raw copy. However ZODB4 goes
    through `.extension` and pickles it:

    https://lab.nexedi.com/nexedi/zodbtools/blob/129afa67/zodbtools/zodbdump.py#L425-453
    https://github.com/zopefoundation/ZODB/blob/4/src/ZODB/BaseStorage.py#L220-L240

    This leads to unpickle-repickle round-trip and different extension being committed on restore:

    ```diff
    diff --git a/1zdump b/2zdump
    index 5033bc1..a3a32aa 100644
    --- a/1zdump
    +++ b/2zdump
    @@ -10,7 +10,7 @@ q^A.
     txn 0285cbac3d0369e6 " "
     user "user0.0"
     description "step 0.0"
    -extension "\x80\x02}q\x01(U\tx-cookieSU\x05RF9IEU\vx-generatorq\x02U\fzodb/py2 (f)u."
    +extension "}q\x01(U\tx-cookieSU\x05RF9IEU\vx-generatorU\fzodb/py2 (f)u."
     obj 0000000000000000 98 sha1:eba252d1984f975ecb636bc1b3a89c953dd20527
    ...
    ```

    What might save us is to somehow in Transaction.extension returns a
    dict-subclass object that is somehow pickled to the exact bytes remembered when
    it was created. However, after briefly checking, I could not find a mechanism
    to do so yet...

@jerome wrote (nexedi/zodbtools!19 (comment 129479))

    @kirr we already have pytest fixtures to test differently depending on whether
    the ZODB version has support for extension_bytes, so what about using it in the
    test and testing restoring the extension bytes version of the dump only for
    ZODB5 ?

@kirr wrote (nexedi/zodbtools!19 (comment 129482))

    @jerome, yes we have this, but I believe we should actually fix zodbrestore to
    be reliable whatever ZODB is used. For ZODB5 it works. For ZODB4-wc2 we can
    adjust ZODB code to use extension_bytes similarly to how ZODB5 does. But
    unpatched ZODB4 is currently out of luck. As it was decided that Nexedi will
    use both ZODB4 and ZODB4-wc2, I think we should fix zodbrestore to work on all
    those versions to be reliable.

    /cc @tomo

@kirr:

-> No universal ZODB4 fix for now (this would require to monkey patch ZODB in
several places), so mark "restore with extension" test as xfail similarly to
how we already do for "dump with extension" test.

This brings -ZODB4 and -ZODB4-wc2 tests back to PASS state.

Even though on ZODB4 extension is restored not bit-to-bit exactly, it is
restored to be the same dictionary equal to what was used to produce the
dump. Not ideal, but still not loosing the information in practice.

One more reason to switch to ZODB5...

In the previous patch we taught object copy handler to report more
details, but it was still incomplete - the error was missing details
about which operation was run - commit, or restore of particular
transaction.

Noting that it can be also noted that other errors reported from that
function lack such context.

-> So fix it universally, at least for zodbcommit for now: set top-level
runctx to topic of what we are doing, and use that runctx when
generating errors. Runctx describes what we are running, and could be
also later used for logging and tracing. That's why it is called runctx
instead of just errctx for "error context".

TODO currently it is only exceptions that we explicitly raise which get
the context. If an exception is raised by something that we call - the
context won't be added. It would be good to later rework error handling
and append such context for any raised error. Defer and
https://lab.nexedi.com/kirr/go123/blob/863c4602/xerr/__init__.py has
something preliminary for this.

The particular error when restoring a missing object copy becomes

    ValueError: /tmp/demo002868462/δ0285cbac75555580/δ.fs: restore 0285cbacb70a3db3 @0285cbacb258bf66: object 0000000000000003: copy from @0285cbac70a3d733: no data

instead of older

    ValueError: /tmp/demo358030847/δ0285cbac75555580/δ.fs: object 0000000000000003: copy from @0285cbac70a3d733: no data

/reviewed-by @jerome
/reviewed-on nexedi/zodbtools!20

When zodbdump input says to copy an object, we first load that object.
However if object does not exist loadBefore raises POSKeyError, and when
object at copied-from revision was deleted loadBefore returns None.

-> Handle that explicitly to provide failure details to the user, so
that instead of cryptic

    === RUN   TestLoad/δstart=0285cbac75555580
    Traceback (most recent call last):
      File "/usr/lib/python2.7/runpy.py", line 174, in _run_module_as_main
        "__main__", fname, loader, pkg_name)
      File "/usr/lib/python2.7/runpy.py", line 72, in _run_code
        exec code in run_globals
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodb.py", line 133, in <module>
        main()
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodb.py", line 129, in main
        return command_module.main(argv)
      File "<decorator-gen-6>", line 2, in main
      File "/home/kirr/src/tools/go/pygolang/golang/__init__.py", line 103, in _
        return f(*argv, **kw)
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodbrestore.py", line 94, in main
        zodbrestore(stor, asbinstream(sys.stdin), _)
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodbrestore.py", line 43, in zodbrestore
        zodbcommit(stor, at, txn)
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodbcommit.py", line 122, in zodbcommit
        _()
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodbcommit.py", line 91, in _
        data, _, _ = stor.loadBefore(obj.oid, p64(u64(obj.copy_from)+1))
    TypeError: 'NoneType' object is not iterable
        xtesting.go:483: /tmp/demo009767458/δ0285cbac75555580/δ.fs: zpyrestore: exit status 1

it fails with something more understandable:

    === RUN   TestLoad/δstart=0285cbac75555580
    Traceback (most recent call last):
      File "/usr/lib/python2.7/runpy.py", line 174, in _run_module_as_main
        "__main__", fname, loader, pkg_name)
      File "/usr/lib/python2.7/runpy.py", line 72, in _run_code
        exec code in run_globals
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodb.py", line 133, in <module>
        main()
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodb.py", line 129, in main
        return command_module.main(argv)
      File "<decorator-gen-6>", line 2, in main
      File "/home/kirr/src/tools/go/pygolang/golang/__init__.py", line 103, in _
        return f(*argv, **kw)
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodbrestore.py", line 94, in main
        zodbrestore(stor, asbinstream(sys.stdin), _)
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodbrestore.py", line 43, in zodbrestore
        zodbcommit(stor, at, txn)
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodbcommit.py", line 129, in zodbcommit
        _()
      File "/home/kirr/src/wendelin/z/zodbtools/zodbtools/zodbcommit.py", line 97, in _
        (stor.getName(), ashex(obj.oid), ashex(obj.copy_from)))
    ValueError: /tmp/demo358030847/δ0285cbac75555580/δ.fs: object 0000000000000003: copy from @0285cbac70a3d733: no data
        xtesting.go:483: /tmp/demo358030847/δ0285cbac75555580/δ.fs: zpyrestore: exit status 1

For the implementation it would be easier to use loadAt
(https://github.com/zopefoundation/ZODB/pull/323), but we don't have
that yet.

/reviewed-by @jerome
/reviewed-on nexedi/zodbtools!20

Suggessted by @jerome here: nexedi/zodbtools!19 (comment 129181)

Co-authored-with: Jérome Perrin <jerome@nexedi.com>
/reviewed-on nexedi/zodbtools!19

Zodbrestore is long-coming counterpart to zodbdump.
Implementation is internally based on reworked zodbcommit.

For FileStorage restored database is verified via test to be bit-to-bit
identical to the original.

For NEO it won't be exactly the case, as NEO does not implement
IStorageRestoreable: there is only tpc_begin(tid=...) but no restore().

/helped-by @jerome
/reviewed-on nexedi/zodbtools!19

This current serial will not be needed on new codepaths to be added to
zodbcommit in the next patch.

-> Move the computation to function to trigger it only from places where
knowing current serial is actually needed.

/reviewed-by @jerome
/reviewed-on nexedi/zodbtools!19

Two-phase commit protocol assumes that after tpc_begin, it will be
either successful tpc_vote + tpc_finish, or tpc_abort. We were not
calling tpc_abort on an error, potentially leaving storage in "commit is
in progress" state on an error.

/reviewed-by @jerome
/reviewed-on nexedi/zodbtools!19

Nexedi stack is dropping support for that old ZODB version - see e.g.

- nexedi/slapos@70d05199
- nexedi/neoppod@3a8f6f03
- nexedi/wendelin.core@0802da2b

Regarding test/gen_testdata.py: even though ZODB4 uses zodbpickle, and
so should be able to load pickles encoded with protocol 3 even on
python2, in practice it does not work so well: ZODB4 tests fail if I set

    --- a/src/ZODB/_compat.py
    +++ b/src/ZODB/_compat.py
    @@ -34,7 +34,7 @@
         HIGHEST_PROTOCOL = cPickle.HIGHEST_PROTOCOL
         IMPORT_MAPPING = {}
         NAME_MAPPING = {}
    -    _protocol = 1
    +    _protocol = 3
         FILESTORAGE_MAGIC = b"FS21"
     else:
         # Python 3.x: can't use stdlib's pickle because

-> so continue to preserve protocol < 3 when generating the test
database for compatibility - now with ZODB4/py2.

/reviewed-by @jerome
/reviewed-on nexedi/zodbtools!18

The patch that provides raw-extension functionality was merged into ZODB 5.6:

https://github.com/zopefoundation/ZODB/commit/2f8cc67a3ba3

So when testing with ZODB5 >= 5.6 the tests will excercise code path
that uses txn.extension_bytes, and when testing with ZODB4 the tests
will excercise code path that work-arounds lack of txn.extension_bytes.

/reviewed-by @jerome
/reviewed-on nexedi/zodbtools!18

Nxdtest[1] is tox-like tool to run tests under Nexedi testing
infrastructure.

[1] https://lab.nexedi.com/nexedi/nxdtest

/reviewed-on nexedi/zodbtools!17

Flushing changes from yet another attempt. Still not completely there yet, but closer.

Reviewed-by: @jerome
Reviewed-on: nexedi/zodbtools!16

ashex gives bytes, whereas reference_tid was str.

The sequence cannot be randomly accessed, e.g.

    In [5]: d = {1:2}

    In [6]: kv = d.keys()

    In [7]: kv
    Out[7]: dict_keys([1])

    In [8]: kv[0]
    ---------------------------------------------------------------------------
    TypeError                                 Traceback (most recent call last)
    <ipython-input-8-643f90e1910b> in <module>()
    ----> 1 kv[0]

    TypeError: 'dict_keys' object is not subscriptable

-> Use list(dict.keys()) in places where we need random access.

Otherwise it breaks with str on py3:

	In [1]: from io import BytesIO

	In [2]: BytesIO("abc")
	---------------------------------------------------------------------------
	TypeError                                 Traceback (most recent call last)
	<ipython-input-2-52a130edd46d> in <module>()
	----> 1 BytesIO("abc")

	TypeError: a bytes-like object is required, not 'str'

Zodbdump format is text-binary and is saved into files opened in binary
mode. -> We have to emit bytes - not strings - into it, since otherwise
on Python3 it would break.

This needs qq support from pygolang[1] to be able to use qq with both
string and bytestring format, e.g. for

	 "hello %s" % qq(name),	and
	b"hello %s" % qq(name)

to give the same output irregardless of whether name is str or bytes.

[1] nexedi/pygolang!1

Zodbdump format is already described as semi text-binary in top-level
zodbdump.py documentation. However zdump() docstring was referring to it
as "text". Fix it and use binary to handle places where zdump is
loaded/saved.

%r has different output for strings and bytes on python3:

	In [1]: a = 'hello'
	In [2]: b = b'hello'

	In [3]: repr(a)
	Out[3]: "'hello'"

	In [4]: repr(b)
	Out[4]: "b'hello'"

-> Use qq whose output is stable irregardless of whether input is string or bytes.

zodbinfo: Provide "head" as command to query DB head; Turn "last_tid" into deprecated alias for head

Similarly to go version: kirr/neo@151d8b79.

Starting with upcoming ZODB 5.5.2 ZODB tries to preserve
`extension_bytes` transaction metadata property in the raw form as it
was stored on disk in the database:

    https://github.com/zopefoundation/ZODB/commit/2f8cc67a

However now when running test/gen_testdata.py with ZODB with that patch (and
gen_testdata.py refuses to work if it detects that ZODB does not properly
supports .extension_bytes property because we want it to be present in the
generated test database [1,2]) it now breaks:

    $ ./gen_testdata.py
    Traceback (most recent call last):
      File "./gen_testdata.py", line 230, in <module>
        main()
      File "./gen_testdata.py", line 224, in main
        gen_testdb("%s.fs" % dbname, zext=zext)
      File "./gen_testdata.py", line 194, in gen_testdb
        stor.tpc_begin(txn)
      File "/home/kirr/src/wendelin/z/ZODB/src/ZODB/BaseStorage.py", line 193, in tpc_begin
        ext = transaction.extension_bytes
    AttributeError: 'Transaction' object has no attribute 'extension_bytes'

The breakage is because, as specified in ZODB interfaces[3,4], storage requires
ZODB.IStorageTransactionMetaData, not transaction.ITransaction instance
gen_testdata.py was using. The script used to work before just by luck.

The fix is to convert transaction instance into storage transaction metadata
object for the place where we talk to storage at raw level.

HOWEVER, when checking regenerated database and its dump I noticed:

ZODB >= 5.4.0 uses pickle protocol 3 on both python2 and python3

    https://github.com/zopefoundation/ZODB/commit/12ee41c4

In other words it saves e.g. OID of an object as pickle binary, which decodes
as bytes on py3 and zodbpickle.binary on py2 when decoding via zodbpickle.
However it will result in *DecodeError* when decoding on py2 with standard
pickle module. The latter means that ZODB3 will _fail_ to load data from test
database, because ZODB3 - contrary to ZODB4 and ZODB5 - uses std pickle module,
not zodbpickle.

We still care about ZODB3 and in particular it is included into
zodbtools test matrix:

    https://lab.nexedi.com/nexedi/zodbtools/blob/7bc0385e/tox.ini#L9-14

so we cannot break it.

-> Temporarily patch ZODB at runtime to make sure it emits data with
older protocol and without using zodbpickle.binary for oid, so that
generated test database could be loaded on ZODB3 as well.

gen_testdata.py now works with latest ZODB, but produces exactly the
same bit-to-bit output as before.

[1] https://lab.nexedi.com/nexedi/zodbtools/blob/7bc0385e/zodbtools/test/gen_testdata.py#L215-217
[2] https://lab.nexedi.com/nexedi/zodbtools/blob/7bc0385e/zodbtools/test/testutil.py#L31-63
[3] https://github.com/zopefoundation/ZODB/blob/5.5.1-35-gb5895a5c2/src/ZODB/interfaces.py#L815-L818
[4] https://github.com/zopefoundation/ZODB/blob/5.5.1-35-gb5895a5c2/src/ZODB/interfaces.py#L538-L575

/reviewed-on !15

-> Use .[test] to refer to them.
https://stackoverflow.com/a/41398850/9456786

/reviewed-by @jerome
/reviewed-on nexedi/zodbtools!14

@jerome, I was trying to make zodbtools work with Python3 and along that road picked some bits of your work from nexedi/zodbtools!12. At present the migration to Python3 is not complete, and even though now I have the answer to how handle strings in both python2/3 in compatible and reasonable way (I can share details if you are interested), I have to put that work on hold for some time and use https://pypi.org/project/pep3134 directly in wcfs tests, since getting all string details right, even after figuring on how to do it, will take time. Anyway the bits presented here should be ready for master and could be merged now. Could you please have a look?

Thanks beforehand,  
Kirill

/reviewed-on nexedi/zodbtools!13

Zodbdump format is mixed text+binary so dumping to unicode stdout won't
work.

Based on patch by Jérome Perrin.

Because on Py3:

        def test_dumpreader():
            in_ = b"""\
        txn 0123456789abcdef " "
        user "my name"
        description "o la-la..."
        extension "zzz123 def"
        obj 0000000000000001 delete
        obj 0000000000000002 from 0123456789abcdee
        obj 0000000000000003 54 adler32:01234567 -
        obj 0000000000000004 4 sha1:9865d483bc5a94f2e30056fc256ed3066af54d04
        ZZZZ
        obj 0000000000000005 9 crc32:52fdeac5
        ABC

        DEF!

        txn 0123456789abcdf0 " "
        user "author2"
        description "zzz"
        extension "qqq"

        """

            r = DumpReader(BytesIO(in_))
            t1 = r.readtxn()
            assert isinstance(t1, Transaction)
    >       assert t1.tid == '0123456789abcdef'.decode('hex')
    E       AttributeError: 'str' object has no attribute 'decode'

    test/test_dump.py:77: AttributeError

Based on patch by Jérome Perrin.

	self = <zodbtools.util.CRC32Hasher object at 0x7f887ae465f8>

	    def __init__(self):
	>       self._h = crc32('')
	E       TypeError: a bytes-like object is required, not 'str'

	util.py:208: TypeError

Based on patch by Jérome Perrin.

	data = 'data1'

	    def sha1(data):
	        m = hashlib.sha1()
	>       m.update(data)
	E       TypeError: Unicode-objects must be encoded before hashing

	zodbtools/util.py:38: TypeError

Based on patch by Jérome Perrin.

	s = b'\x03\xc4\x85v\x00\x00\x00\x00'

	    def ashex(s):
	>       return s.encode('hex')
	E       AttributeError: 'bytes' object has no attribute 'encode'

	zodbtools/util.py:29: AttributeError

s.encode('hex') used to work on Py2 but fails on Py3:

	In [1]: s = "abc"

	In [2]: b = b"def"

	In [3]: s.encode('hex')
	---------------------------------------------------------------------------
	LookupError                               Traceback (most recent call last)
	<ipython-input-3-75ae843597fe> in <module>()
	----> 1 s.encode('hex')

	LookupError: 'hex' is not a text encoding; use codecs.encode() to handle arbitrary codecs

	In [4]: b.encode('hex')
	---------------------------------------------------------------------------
	AttributeError                            Traceback (most recent call last)
	<ipython-input-4-ec2fccff20bc> in <module>()
	----> 1 b.encode('hex')

	AttributeError: 'bytes' object has no attribute 'encode'

	In [5]: import codecs

	In [6]: codecs.encode(b, 'hex')
	Out[6]: b'646566'

	In [7]: codecs.encode(s, 'hex')
	---------------------------------------------------------------------------
	TypeError                                 Traceback (most recent call last)
	/usr/lib/python3.7/encodings/hex_codec.py in hex_encode(input, errors)
	     14     assert errors == 'strict'
	---> 15     return (binascii.b2a_hex(input), len(input))
	     16

	TypeError: a bytes-like object is required, not 'str'

	The above exception was the direct cause of the following exception:

	TypeError                                 Traceback (most recent call last)
	<ipython-input-7-7fcb16cead4f> in <module>()
	----> 1 codecs.encode(s, 'hex')

	TypeError: encoding with 'hex' codec failed (TypeError: a bytes-like object is required, not 'str')

After the patch it works with bytes and raises for str.
Fromhex does not need to be changed - it already uses codecs.decode way as
originally added in dd959b28 (zodbdump += DumpReader - to read/parse zodbdump
stream).

Based on patch by Jérome Perrin.

There is no cStringIO on Python3:

	test_dump.py:26: in <module>
	    from cStringIO import StringIO
	E   ModuleNotFoundError: No module named 'cStringIO'

Based on patch by Jérome Perrin.

This makes zodb command driver tests added in the previous patch to pass
on both python2 and python3.

----

kirr: factor running `zodb ...` into zodbrun + add test for `zodb -h`.

Added test currently passes on py2, but fails on py3:

	out = <_io.TextIOWrapper encoding='UTF-8'>

	    def usage(out):
	        print("""\
	    Zodb is a tool for managing ZODB databases.

	    Usage:

	        zodb command [arguments]

	    The commands are:
	    """, file=out)

	        cmdv = command_dict.keys()
	>       cmdv.sort()
	E       AttributeError: 'dict_keys' object has no attribute 'sort'

	zodbtools/zodb.py:55: AttributeError

It will be fixed in the next patch.

----

kirr: use loggging as log and keep emitting warnings on one line.

this silents a warning about \w being unknown escape sequence

----

kirr: preserved _obj_re definition to be on 1 line.

until https://github.com/zopefoundation/ZODB/pull/183 gets merged, let's
run also the tests for this, since we have support for this extension.

also simplify a bit definition as ZODB is common in all versions

----

kirr:

- cover only last 2 py3 releases: 3.6 and 3.7 for now (3.8 is not yet
  released)
- separate ZODB3 as it supports only python2.

Py3 tests are failing for now and we'll be getting them to pass
incrementally - step by step.

Using dateparser to support absolute and relative dates in natural language.

/reviewed-by @kirr
/reviewed-on nexedi/zodbtools!8