Rendering a cascade or trace Diagram fails on PostgreSQL: quoted table names reach the DOT source unescaped

Ouverte
#1,692 0 commentaires 0 réactions 0 personnes assignées Voir sur GitHub

Les mainteneurs répondent en général sous 6 jours

Personne n'a encore pris cette issue.

Évaluation

Difficulté
3/5
Temps estimé
1-2 jours
Accessibilité débutants
70/100
Type d'issue
Bug
Clarté
Clairement spécifiée
Activité
Active
Stack technique
postgresql, python

Piste de recherche

Start in diagram.py: _make_graph (~1309) uses full table names as node ids, _encapsulate_node_names (~1550) adds quotes, and the plain-label branch (~1756) emits label="schema"."table". Read adapter.split_full_table_name, use the unquoted schema.table for unresolved ids and labels, and keep the strip('"') lookups in make_dot (1675, 1809) consistent so schema clusters and master-part groups appear. Re-run the PostgreSQL repro; done when _repr_svg_(), save("x.svg") and save("x.png") render with nodes and edges intact.

Rédigé par le modèle d'indexation à partir du texte de l'issue.

Description

bug

Summary

On PostgreSQL, rendering a dj.Diagram whose nodes are shown by their full table names fails when any shown table inherits its whole primary key through foreign keys (a 1:1 extension, or a typical computed table, drawn without an underline). _repr_svg_(), save("x.svg") and save("x.png") raise:

AssertionError: "dot" with args ['-Tsvg', '<tmpfile>'] returned code: 1

pydot prints Graphviz's message to stdout: Error: <tmpfile>: syntax error in line N near '.'. make_dot() and make_mermaid() themselves do not raise. Diagram.cascade(...).counts() and iterating the cascade work.

Reproduced on master (f6299387) with PostgreSQL 15, pydot 4.0.1 and Graphviz 15.1.0. Every case renders on MySQL.

When it happens

  • Nodes keep their full table names. _make_graph names each node by its class when lookup_class_name finds it in the diagram's context and by its full table name otherwise (diagram.py:1309). Diagram.cascade and Diagram.trace set an empty context (diagram.py:504, 589), so their nodes are always full names. A Diagram built with a context that lacks the table classes does the same.
  • On PostgreSQL, full names are "schema"."table". _encapsulate_node_names wraps them in another pair of double quotes (diagram.py:1550), and pydot passes them through unescaped.
  • A table that adds no key attribute of its own gets a plain label (diagram.py:1756): label="schema"."table". The bare . is the syntax error.

When no shown table has a plain label, the render succeeds but is wrong: Graphviz splits each id into several nodes ("", the schema name, ".", the table name), edges collapse, and no schema cluster is drawn.

A master-part group is not the trigger: a cascade over a master and its part where every table adds its own key attribute renders without error.

Reproduce

import datajoint as dj
# PostgreSQL connection: dj.config["database.backend"] = "postgresql", plus host, user, password
schema = dj.Schema("render_repro")

@schema
class Session(dj.Manual):
    definition = """
    session_id : int32
    """

    class Trial(dj.Part):
        definition = """
        -> master
        trial_id : int32
        """

@schema
class ProcessedData(dj.Manual):
    definition = """
    -> Session.Trial
    ---
    value : float64
    """

Session.insert1({"session_id": 1})
Session.Trial.insert1({"session_id": 1, "trial_id": 1})
ProcessedData.insert1({"session_id": 1, "trial_id": 1, "value": 1.0})

cascade = dj.Diagram.cascade(Session & {"session_id": 1})
print(cascade.counts())   # works on both backends
cascade._repr_svg_()      # MySQL: renders. PostgreSQL: AssertionError from pydot

Fix

Use a backend-neutral display name for unresolved nodes, for example the unquoted schema.table from adapter.split_full_table_name, for both the DOT node id and its label, so no raw " reaches pydot. Alternatively escape the embedded quotes in both. Keep the strip('"') lookups in make_dot (diagram.py:1675, 1809) consistent with the new ids, so schema clusters and master-part groups appear on PostgreSQL.

PostgreSQL is planned to become the default backend in 2.4, and the delete docs suggest rendering the cascade as a preview, so this should be fixed in 2.4.0.

Langage dominant
Python
Étoiles
198
Forks
103
Merge moyen
6 j 1 h
PR mergées (30 j)
10

Préparer son environnement

Par où commencer

  1. Lisez l'issue en entier, puis le guide de contribution du projet.
  2. Signalez en commentaire que vous la prenez — cela évite que deux personnes fassent le même travail.
  3. Forkez le dépôt et travaillez sur une branche.
  4. Ouvrez une pull request qui référence le numéro de l'issue.

Autres issues de datajoint/datajoint-python

Toutes les issues de datajoint/datajoint-python

Issues similaires

Plus d'issues Python

Recevez les nouvelles issues par e-mail

Un résumé court des issues GitHub adaptées aux débutants.