CODEWITHSUNDEEP
sqlalchemyhexvarbinarymemoryview
meisterkaio
meisterkaio

Reputation: 5

Reading VARBINARY via SQLAlchemy

I am working on a base-datatable with a VARBINARY variable. Now I want to read the table via SQLAlchemy into a pandas dataframe. Going the usual way

df = pandas.read_sql_query("select key from xxx", engine)

I get an uninterpretable memoryview as data type. I can convert this via lambda function

df.key.apply(lambda x: x.tobytes().hex())

into the desired readable format. But I would like to know if the casting can also be placed directly into the pandas.read_sql_query()-statement:

Many greetings and best thanks

Upvotes: 0

Views: 319

Answers (1)

Raul Chiarella
Raul Chiarella

Reputation: 722

I am not sure if this will help you but, thanks to @JGFMK response, I was able to come up with something similar on my program:

  1. Defined the property using Mapped to convert the value that has a DATA_TYPE from the SQL Server.
  2. Used a decode method on the property, according to the collation my SQL Server is using, going to Properties > General on my Database.

This was the result:

    # `db_sql.VARBINARY` informs the `DATA_TYPE` from the table field.
    # Mapped does the conversion from table type to python type.
    order_description_binary: Mapped[bytes] = db_sql.Column('TJ_OBSERVA', db_sql.VARBINARY, nullable=True, default=None)

    @property
    def ordem_description(self):
        if self.order_description_binary is not None:
            try:
                description_text = self.order_description_binary .decode('latin1')
                description_text = description_text.replace('\x00', '')
                return description_text
            except UnicodeDecodeError as e:
                print(f"Error decoding order_description_binary: {e}")
                return None
        else:
            return None

So, in your case you can try using decode alongside replace on the specific data on your dataframe that has VARBINARY type.

Example:

df = pandas.read_sql_query("select key from xxx", engine)
df.specific_data.decode('latin1').replace('\x00', '')

I believe something like that would work.

Upvotes: 0

Related Questions