id	summary	reporter	owner	description	type	status	component	version	severity	resolution	keywords	cc	stage	has_patch	needs_docs	needs_tests	needs_better_patch	easy	ui_ux
31846	MySQL inspectdb doesn't handle composite keys	François Poulain		"Hi,

I plugged django on a crappy drupal DB using inspectdb.py. It is intended to be read only. I got models like

{{{
class FeedapiStat(models.Model):
    id = models.PositiveIntegerField()
    type = models.CharField(max_length=64)
    timestamp = models.IntegerField()
    time = models.CharField(max_length=20)
    value = models.IntegerField()
}}}

And I have been facing with two issues:

* drupal.FeedapiStat: (models.E004) 'id' can only be used as a field name if the field also sets 'primary_key=True'.
* Then I renamed the field and faced with: (models.E007) Field 'id_field' has column name 'id' that is used by another field. HINT: Specify a 'db_column' for the field. This is because the model has no pk, so django automatically try to add one.

I also got some models with a primary key and an id field which is not a pk. 

Since there is no way to tell that a model should not have pk, I modified inspectdb as follows:

{{{
--- ./venv/lib/python3.7/site-packages/django/core/management/commands/inspectdb.py	2020-08-01 10:28:06.155225992 +0200
+++ ./base/management/commands/inspectdb.py	2020-08-01 10:35:06.677586957 +0200
@@ -101,8 +101,10 @@
                     column_name = row.name
                     is_relation = column_name in relations
 
+                    is_pk = column_name == primary_key_column
+                    has_pk = primary_key_column
                     att_name, params, notes = self.normalize_col_name(
-                        column_name, used_column_names, is_relation)
+                        column_name, used_column_names, is_relation, is_pk, has_pk)
                     extra_params.update(params)
                     comment_notes.extend(notes)
 
@@ -110,7 +112,7 @@
                     column_to_field_name[column_name] = att_name
 
                     # Add primary_key and unique, if necessary.
-                    if column_name == primary_key_column:
+                    if is_pk or (column_name == 'id' and not has_pk):
                         extra_params['primary_key'] = True
                     elif column_name in unique_columns:
                         extra_params['unique'] = True
@@ -173,7 +175,7 @@
                 for meta_line in self.get_meta(table_name, constraints, column_to_field_name, is_view, is_partition):
                     yield meta_line
 
-    def normalize_col_name(self, col_name, used_column_names, is_relation):
+    def normalize_col_name(self, col_name, used_column_names, is_relation, is_pk, has_pk):
         """"""
         Modify the column name to make it Python-compatible as a field name
         """"""
@@ -213,6 +215,10 @@
             new_name += '_field'
             field_notes.append('Field renamed because it was a Python reserved word.')
 
+        if new_name == 'id' and not is_pk and has_pk:
+            new_name += '_field'
+            field_notes.append(""Field renamed because 'id' can only be used as a field name if the field also sets 'primary_key=True'"")
+
         if new_name[0].isdigit():
             new_name = 'number_%s' % new_name
             field_notes.append(""Field renamed because it wasn't a valid Python identifier."")
}}}

Not sure this is a bug but adding this feature makes inspectdb's behavior more convenient, I guess, because the models.E007 error is not easily understandable and didn't helped me to find the root of the issue.

Moreover, when a model has nor pk neither id field, a specific warning could be raised, since the automatic id creation will mismatch with the db. Maybe, a model instanciated with {{{Meta: managed = False}}} and without pk should always raise a check warning.

How do you think about it?"	New feature	closed	Core (Management commands)	3.0	Normal	duplicate		Adam Johnson	Accepted	0	0	0	0	0	0