WeakValueDictionary.setdefault() and WeakValueDictionary.pop() will sometimes return None in multithreaded environments, as explained here http://bugs.python.org/issue19542. The patch is pending review.

I've been able to reproduce the problem by running the test case x.py that's linked in the bug report using both python 2.7.9 and python 3.4.3.

I have also encountered the problem during development of a Flask app that uses py2neo. This app is run on Apache using mod_wsgi-express (prefork, 1 process, 5 threads).

Anyhow, since Node.cache, Rel.cache, and Relationship.cache all inherit from WeakValueDictionary and have methods that call setdefault(), I've been seeing NoneType errors in my Flask app.

For instance, this snippet in Rel#hydrate() in core.py:

        if inst is None:
            new_inst = cls()
            new_inst.__stale.update({"properties"})
            inst = cls.cache.setdefault(self, new_inst)

        cls.cache[self] = inst

Will intermittently produce:

    File "/Users/fallonchen/.virtualenvs/ul06/lib/python2.7/site-packages/py2neo/core.py", line 912, in match
    for result in results:
    File "/Users/fallonchen/.virtualenvs/ul06/lib/python2.7/site-packages/py2neo/cypher/core.py", line 453, in next
    return self.__next__()
    File "/Users/fallonchen/.virtualenvs/ul06/lib/python2.7/site-packages/py2neo/cypher/core.py", line 450, in __next__
    return next(self.__response_item)
    File "/Users/fallonchen/.virtualenvs/ul06/lib/python2.7/site-packages/py2neo/cypher/core.py", line 443, in __response_iterator
    yield producer.produce(self.graph.hydrate(assembled(record_data)))
    File "/Users/fallonchen/.virtualenvs/ul06/lib/python2.7/site-packages/py2neo/core.py", line 839, in hydrate
    return type(data)(map(self.hydrate, data))
    File "/Users/fallonchen/.virtualenvs/ul06/lib/python2.7/site-packages/py2neo/core.py", line 806, in hydrate
    return Relationship.hydrate(data)
    File "/Users/fallonchen/.virtualenvs/ul06/lib/python2.7/site-packages/py2neo/core.py", line 2421, in hydrate
    Rel.hydrate(data),
    File "/Users/fallonchen/.virtualenvs/ul06/lib/python2.7/site-packages/py2neo/core.py", line 1769, in hydrate
    cls.cache[self] = inst
    File "/usr/local/Cellar/python/2.7.8_1/Frameworks/Python.framework/Versions/2.7/lib/python2.7/weakref.py", line 103, in __setitem__
    self.data[key] = KeyedRef(value, self._remove, key)
    File "/usr/local/Cellar/python/2.7.8_1/Frameworks/Python.framework/Versions/2.7/lib/python2.7/weakref.py", line 267, in __new__
    self = ref.__new__(type, ob, callback)
    TypeError: cannot create weak reference to 'NoneType' object

I've found that you can get around the problem by replacing setdefault() with:

        if self not in cls.cache:
             cls.cache[self] = new_inst
         inst = cls.cache[self]

Is this a good approach? I haven't looked too deeply into whether or not pop() is called, but I assume that it would also have to be patched.

Using neo4j 4.0.0

Using py2neo v4

g.nodes.match(U, _id=3125349375).first()

yields:

ClientError: SyntaxError: The old parameter syntax `{param}` is no longer supported. 
Please use `$param` instead (line 1, column 30 (offset: 29))
"MATCH (_:USER) WHERE _._id = {1} RETURN _"

@nigelsmall - Using version 1.6.4, neo4j 2.1.3 and python 3.3, when I try the following

graph_db = neo4j.GraphDatabaseService("http://localhost:7474/db/data/")
n = graph_db.node(100)
n["username"]="Nigel"

I get:

AttributeError("'Node' object has no attribute 'metadata'",)

could this be a bug with using a later version of Neo4j?

It's obvious that you are still working on this project @technige, but it's kind of disappointing that are not many responses to the open issues. Perhaps many of the issues can be closed as simple misunderstandings. For myself, I would like to contribute but I need some guidance. I also learn a lot about a project by responses and activity in issues. I think more activity in the issues could help reduce the bus factor and be an overall benefit to the project. This is just a suggestion, as I enjoy working with this project and want to see it continue to improve

Hi Nigel,

We have been porting neomodel to py2neo 1.6, we have come across one issue. The individual http response status codes seem to be no longer accessible in a batch response.

I am aware this was never part of your public API, however it was used internally by neomodel to aid it safely simulating the create_or_fail index feature on neo4j versions where its not available (such as 1.8 used on heroku).

Setting a key, value in an index returns 200 if it already existed and 201 if its new entry. As these codes are no longer available in py2neo / httpstream we can't capture and report this condition: https://github.com/robinedwards/neomodel/blob/master/neomodel/exception.py#L13

Do you know if theres any way I can monkey patch / subclass the current batch system to return them?

When i run the program in a timed task, this problem occurred. I use flask_apscheduler and py2neo, hoping to update the data in an hour. But this error will happen every once in a while. I don't know how this error is caused. Who knows can tell me what is going on. Thank you very much!

I'm aware that 1.6 is not yet final but I have been testing it in order to use labels and the new schema indexes in a project I am working on. In this project batches of 20-1000 statements are sent and tests are taking approx. 14 times longer on 1.6 than on 1.5, even without assigning any labels (straight swap of the library with same app code). It seems as though there is a delay between streamed requests ending and starting. Is this something you are aware of, if so then cool, but I just wanted to point out my findings in case not.

Hello All, Initially I was using a py2neo's version 2021.1.5 within my project and the database module was getting successfully imported in the code. But, when I have updated the py2neo version to the latest one i.e. 2021.2.0 and then I was trying to import the database module from the py2neo I was getting an OverFlow error mentioning 'OverflowError: mktime argument out of range'. May I know the reason for the same ? Why the database module was getting successfully imported with the older version of py2neo? Thank you in advance.

I had an issue while installing a package of mine (pyramid application), which depended on py2neo.

Issue:

in py2neo setup.py file we have: prompt_toolkit==1.0.15

however, py2neo also depends on (found in setup.py): jupyter_console

in jupyter_console setup.py file we have:

prompt_toolkit>=2.0.0,<2.1.0

This can lead to potential conflict, which will not be explicitly shown when you install py2neo. However, when you use something like pyramid, which utilizes pkg_resources, it can lead to: pkg_resources.ContextualVersionConflict

The OGM would be more useful if some hook would be provided so that the user has more control on how objects are saved and loaded.

Example use case: I'd like to save an object and have the object's class name as a label. Since the store.save will only store the object's attrs as node properties I have to resort to 1) save object 2) reload object 3) set object.node.labels. 4) save object again. Yuck.

PS1 the store.save method seems to use a deprecated function set_properties. PS2 it seems that passing a newly created node (see use case above; I wanted to pass the label) as a kwarg in the store.save(obj, node=mynewnode) does not work. I get a 500 server error. Did not explore further.

For some reason tests are broken with Neo4j 2.1.7, even though they are passing with Neo4j 2.1.6 and 2.2.0-M03.

Looks related to Cypher transactions not returning expected uris, but I don't see any reasons for this in Neo4j release notes.

If I can help with this, please give me some hint.

py2neo version: 2021.2.3 python version: 3.9.15

I created a database with nodes that include coordinates and am able to retrieve them using the basic GUI with this query:

match (n) return n limit 1;

However, when I attempt the same thing using py2neo like this:

graph.run("match (n) return n limit 1;")

I get the following error:

TypeError: Cypher literal values of type interchange.geo.WGS84Point are not supported.

Are there plans to include WGS84Point support? If not, is there perhaps a work-around?

Traceback (most recent call last): File "/Users/KGQA_Version.Amme/QASystemOnMedicalKG-AmmeRevision/build_medicalgraph.py", line 346, in handler.create_graphnodes(); File "/Users/KGQA_Version.Amme/QASystemOnMedicalKG-AmmeRevision/build_medicalgraph.py", line 250, in create_graphnodes self.create_diseases_nodes(disease_infos) File "/Users/KGQA_Version.Amme/QASystemOnMedicalKG-AmmeRevision/build_medicalgraph.py", line 237, in create_diseases_nodes self.g.create(node) File "/Users/anaconda3/envs/torch107/lib/python3.7/site-packages/py2neo/database.py", line 362, in create with self.begin() as tx: File "/Users/anaconda3/envs/torch107/lib/python3.7/site-packages/py2neo/database.py", line 353, in begin return Transaction(self, autocommit) File "/Users/anaconda3/envs/torch107/lib/python3.7/site-packages/py2neo/database.py", line 781, in init self.transaction = self.connector.begin() File "/Users/anaconda3/envs/torch107/lib/python3.7/site-packages/py2neo/internal/connectors.py", line 398, in begin raise RuntimeError("Can't begin a new transaction") RuntimeError: Can't begin a new transaction

Process finished with exit code 1

how can I solve it ? thank you!

Hi everyone, We are working with the new version of neo4j (version 5). Are you working in the adaptation of py2neo package for this new version?? Thanks

I use the latest py2neo version and in the server logs I can see this error message keeps appearing (not always). It affects the users of the application sometimes. Failed to read message

I tried to debug the issue. I am using bolt connection inside an azure app service. This happens when the graph.run(query).data() is executed and after few minutes (usually 3 - 4 minutes) this throws the'Failed to read message' exception. It seems this is thrown inside the Wire read. I have seen a similar issue reported here #844 Any solution or workaround is highly appreciated.

Issue

Description

from py2neo import Graph, Node
graph = Graph(NEO4J_URI, user=NEO4J_USER, password=NEO4J_PASSWORD)
node_a = Node("Node", name="A")
node_b = Node("Node", name="B")

graph_tx = graph.begin()
for node in (node_a, node_b):
    graph_tx.create(node)
graph.commit(graph_tx)

relationship_1 = Relationship(node_a, "REL", node_b, id_overwritten=6)
relationship_2 = Relationship(node_a, "REL", node_b, id_overwritten=7)

graph_tx = graph.begin()
for node in (node_a, node_b):
    graph_tx.create(node)
graph.commit(graph_tx)

graph_tx = graph.begin()
for relationship in (relationship_1, relationship_2):
    graph_tx.create(relationship)
graph.commit(graph_tx)

Expected behavior

Given that relationship_2 has a different id_overwritten, a new relationship should be created.

(A) -[REL {"id_overwritten": 6}]-> (B)
(A) -[REL {"id_overwritten": 7}]-> (B)

Actual behavior

relationship_1 is replaced with relationship_2

(A) -[REL {"id_overwritten": 7}]-> (B)

Cause

This is due to the fact that Subgraph.__db_create__ has no option for CREATEing instead of MERGEing

https://github.com/py2neo-org/py2neo/blob/2e46bbf4d622f53282e796ffc521fc4bc6d0b60d/py2neo/data.py#L209

>>> print(pq[0])

UNWIND $data AS r
MATCH (a) WHERE id(a) = r[0]
MATCH (b) WHERE id(b) = r[2]
MERGE (a)-[_:REL]->(b)
SET _ += r[1]

Workaround

...
pq = unwind_merge_relationships_query(data, r_type)

import re

# Pardon my regex
pq = (
    re.sub(
        (
            r"(.*)?"
            r"(MERGE )(\(a\)-\[_:\w+\]->\(b\))"
            r"(.*)"
        ),
        (
            r"\g<1>"
            r"CREATE "
            r"\g<3>"
            r"\g<4>"
        ),
        pq[0],
    ),
    *pq[1:],
)
...

so that

>>> print(pq[0])

UNWIND $data AS r
MATCH (a) WHERE id(a) = r[0]
MATCH (b) WHERE id(b) = r[2]
CREATE (a)-[_:REL]->(b)
SET _ += r[1]

What happened？

There are 1 security vulnerabilities found in pygments 2.0.0

• CVE-2021-20270

What did I do？

Upgrade pygments from 2.0.0 to 2.7.4 for vulnerability fix

What did you expect to happen？

Ideally, no insecure libs should be used.

The specification of the pull request

PR Specification from OSCS