SQL select join: is it possible to prefix all columns as 'prefix.*'?

It seems the answer to your question is no, however one hack you can use is to assign a dummy column to separate each new table. This works especially well if you're looping through a result set for a list of columns in a scripting language such as Python or PHP.

SELECT '' as table1_dummy, table1.*, '' as table2_dummy, table2.*, '' as table3_dummy, table3.* FROM table1
JOIN table2 ON table2.table1id = table1.id
JOIN table3 ON table3.table1id = table1.id

I realize this doesn't answer your question exactly, but if you're a coder this is a great way to separate tables with duplicate column names. Hope this helps somebody.

I see two possible situations here. First, you want to know if there is a SQL standard for this, that you can use in general regardless of the database. No, there is not. Second, you want to know with regard to a specific dbms product. Then you need to identify it. But I imagine the most likely answer is that you'll get back something like "a.id, b.id" since that's how you'd need to identify the columns in your SQL expression. And the easiest way to find out what the default is, is just to submit such a query and see what you get back. If you want to specify what prefix comes before the dot, you can use "SELECT * FROM a AS my_alias", for instance.

The only database I know that does this is SQLite, depending on the settings you configure with PRAGMA full_column_names and PRAGMA short_column_names. See http://www.sqlite.org/pragma.html

Otherwise all I can recommend is to fetch columns in a result set by ordinal position rather than by column name, if it's too much trouble for you to type the names of the columns in your query.

This is a good example of why it's bad practice to use SELECT * -- because eventually you'll have a need to type out all the column names anyway.

I understand the need to support columns that may change name or position, but using wildcards makes that harder, not easier.

I am in kind of the same boat as OP - I have dozens of fields from 3 different tables that I'm joining, some of which have the same name(ie. id, name, etc). I don't want to list each field, so my solution was to alias those fields that shared a name and use select * for those that have a unique name.

For example :

table a : id, name, field1, field2 ...

table b : id, name, field3, field4 ...

select a.id as aID, a.name as aName, a. * , b.id as bID, b.name as bName, b. * .....

When accessing the results I us the aliased names for these fields and ignore the "original" names.

Maybe not the best solution but it works for me....i'm use mysql

In postgres, I use the json functions to instead return json objects.... then, after querying, I json_decode the fields with a _json suffix.

IE:

select row_to_json(tab1.*),tab1_json, row_to_json(tab2.*) tab2_json 
 from tab1
 join tab2 on tab2.t1id=tab1.id

then in PHP (or any other language), I loop through the returned columns and json_decode() them if they have the "_json" suffix (also removing the suffix. In the end, I get an object called "tab1" that includes all tab1 fields, and another called "tab2" that includes all tab2 fields.

DIfferent database products will give you different answers; but you're setting yourself up for hurt if you carry this very far. You're far better off choosing the columns you want, and giving them your own aliases so the identity of each column is crystal-clear, and you can tell them apart in the results.

There is no SQL standard for this.

However With code generation (either on demand as the tables are created or altered or at runtime), you can do this quite easily:

CREATE TABLE [dbo].[stackoverflow_329931_a](
    [id] [int] IDENTITY(1,1) NOT NULL,
    [col2] [nchar](10) NULL,
    [col3] [nchar](10) NULL,
    [col4] [nchar](10) NULL,
 CONSTRAINT [PK_stackoverflow_329931_a] PRIMARY KEY CLUSTERED 
(
    [id] ASC
)WITH (PAD_INDEX  = OFF, STATISTICS_NORECOMPUTE  = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS  = ON, ALLOW_PAGE_LOCKS  = ON) ON [PRIMARY]
) ON [PRIMARY]

CREATE TABLE [dbo].[stackoverflow_329931_b](
    [id] [int] IDENTITY(1,1) NOT NULL,
    [col2] [nchar](10) NULL,
    [col3] [nchar](10) NULL,
    [col4] [nchar](10) NULL,
 CONSTRAINT [PK_stackoverflow_329931_b] PRIMARY KEY CLUSTERED 
(
    [id] ASC
)WITH (PAD_INDEX  = OFF, STATISTICS_NORECOMPUTE  = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS  = ON, ALLOW_PAGE_LOCKS  = ON) ON [PRIMARY]
) ON [PRIMARY]

DECLARE @table1_name AS varchar(255)
DECLARE @table1_prefix AS varchar(255)
DECLARE @table2_name AS varchar(255)
DECLARE @table2_prefix AS varchar(255)
DECLARE @join_condition AS varchar(255)
SET @table1_name = 'stackoverflow_329931_a'
SET @table1_prefix = 'a_'
SET @table2_name = 'stackoverflow_329931_b'
SET @table2_prefix = 'b_'
SET @join_condition = 'a.[id] = b.[id]'

DECLARE @CRLF AS varchar(2)
SET @CRLF = CHAR(13) + CHAR(10)

DECLARE @a_columnlist AS varchar(MAX)
DECLARE @b_columnlist AS varchar(MAX)
DECLARE @sql AS varchar(MAX)

SELECT @a_columnlist = COALESCE(@a_columnlist + @CRLF + ',', '') + 'a.[' + COLUMN_NAME + '] AS [' + @table1_prefix + COLUMN_NAME + ']'
FROM INFORMATION_SCHEMA.COLUMNS
WHERE TABLE_NAME = @table1_name
ORDER BY ORDINAL_POSITION

SELECT @b_columnlist = COALESCE(@b_columnlist + @CRLF + ',', '') + 'b.[' + COLUMN_NAME + '] AS [' + @table2_prefix + COLUMN_NAME + ']'
FROM INFORMATION_SCHEMA.COLUMNS
WHERE TABLE_NAME = @table2_name
ORDER BY ORDINAL_POSITION

SET @sql = 'SELECT ' + @a_columnlist + '
,' + @b_columnlist + '
FROM [' + @table1_name + '] AS a
INNER JOIN [' + @table2_name + '] AS b
ON (' + @join_condition + ')'

PRINT @sql
-- EXEC (@sql)

Or you could use Red Gate SQL Refactor or SQL Prompt, which expands your SELECT * into column lists with a click of the Tab button

so in your case, if you type in SELECT * FROM A JOIN B ... Go to the end of *, Tab button, voila! you'll see SELECT A.column1, A.column2, .... , B.column1, B.column2 FROM A JOIN B

It's not free though

select * usually makes for bad code, as new columns tend to get added or order of columns change in tables quite frequently which usually breaks select * in a very subtle ways. So listing out columns is the right solution.

As to how to do your query, not sure about mysql but in sqlserver you could select column names from syscolumns and dynamically build the select clause.

There are two ways I can think of to make this happen in a reusable way. One is to rename all of your columns with a prefix for the table they have come from. I have seen this many times, but I really don't like it. I find that it's redundant, causes a lot of typing, and you can always use aliases when you need to cover the case of a column name having an unclear origin.

The other way, which I would recommend you do in your situation if you are committed to seeing this through, is to create views for each table that alias the table names. Then you join against those views, rather than the tables. That way, you are free to use * if you wish, free to use the original tables with original column names if you wish, and it also makes writing any subsequent queries easier because you have already done the renaming work in the views.

Finally, I am not clear why you need to know which table each of the columns came from. Does this matter? Ultimately what matters is the data they contain. Whether UserID came from the User table or the UserQuestion table doesn't really matter. It matters, of course, when you need to update it, but at that point you should already know your schema well enough to determine that.

If concerned about schema changes this might work for you: 1. Run a 'DESCRIBE table' query on all tables involved. 2. Use the returned field names to dynamically construct a string of column names prefixed with your chosen alias.

PHP 7.2 + MySQL/Mariadb

MySQL will send you multiple fields with the same name. Even in the terminal client. But if you want an associative array, you'll have to make the keys yourself.

Thanks to @axelbrz for the original. I've ported it to newer php and cleaned it up a little:

function mysqli_rows_with_columns($link, $query) {
    $result = mysqli_query($link, $query);
    if (!$result) {
        return mysqli_error($link);
    }
    $field_count = mysqli_num_fields($result);
    $fields = array();
    for ($i = 0; $i < $field_count; $i++) {
        $field = mysqli_fetch_field_direct($result, $i);
        $fields[] = $field->table . '.' . $field->name; # changed by AS
        #$fields[] = $field->orgtable . '.' . $field->orgname; # actual table/field names
    }
    $rows = array();
    while ($row = mysqli_fetch_row($result)) {
        $new_row = array();
        for ($i = 0; $i < $field_count; $i++) {
            $new_row[$fields[$i]] = $row[$i];
        }
        $rows[] = $new_row;
    }
    mysqli_free_result($result);
    return $rows;
}

$link = mysqli_connect('localhost', 'fixme', 'fixme', 'fixme');
print_r(mysqli_rows_with_columns($link, 'select foo.*, bar.* from foo, bar'));

I implemented a solution based upon the answer suggesting using dummy or sentinel columns in node. You would use it by generating SQL like:

select 
    s.*
  , '' as _prefix__creator_
  , u.*
  , '' as _prefix__speaker_
  , p.*
from statements s 
  left join users u on s.creator_user_id = u.user_id
  left join persons p on s.speaker_person_id = p.person_id

And then post-processing the row you get back from your database driver like addPrefixes(row).

Implementation (based upon the fields/rows returned by my driver, but should be easy to change for other DB drivers):

const PREFIX_INDICATOR = '_prefix__'
const STOP_PREFIX_INDICATOR = '_stop_prefix'

/** Adds a <prefix> to all properties that follow a property with the name: PREFIX_INDICATOR<prefix> */
function addPrefixes(fields, row) {
  let prefix = null
  for (const field of fields) {
    const key = field.name
    if (key.startsWith(PREFIX_INDICATOR)) {
      if (row[key] !== '') {
        throw new Error(`PREFIX_INDICATOR ${PREFIX_INDICATOR} must not appear with a value, but had value: ${row[key]}`)
      }
      prefix = key.substr(PREFIX_INDICATOR.length)
      delete row[key]
    } else if (key === STOP_PREFIX_INDICATOR) {
      if (row[key] !== '') {
        throw new Error(`STOP_PREFIX_INDICATOR ${STOP_PREFIX_INDICATOR} must not appear with a value, but had value: ${row[key]}`)
      }
      prefix = null
      delete row[key]
    } else if (prefix) {
      const prefixedKey = prefix + key
      row[prefixedKey] = row[key]
      delete row[key]
    }
  }
  return row
}

Test:

const {
  addPrefixes,
  PREFIX_INDICATOR,
  STOP_PREFIX_INDICATOR,
} = require('./BaseDao')

describe('addPrefixes', () => {
  test('adds prefixes', () => {
    const fields = [
      {name: 'id'},
      {name: PREFIX_INDICATOR + 'my_prefix_'},
      {name: 'foo'},
      {name: STOP_PREFIX_INDICATOR},
      {name: 'baz'},
    ]
    const row = {
      id: 1,
      [PREFIX_INDICATOR + 'my_prefix_']: '',
      foo: 'bar',
      [STOP_PREFIX_INDICATOR]: '',
      baz: 'spaz'
    }
    const expected = {
      id: 1,
      my_prefix_foo: 'bar',
      baz: 'spaz',
    }
    expect(addPrefixes(fields, row)).toEqual(expected)
  })
})

What I do is use Excel to concatenate the procedure. For instance, first I select * and get all of the columns, paste them in Excel. Then write out the code I need to surround the column. Say i needed to ad prev to a bunch of columns. I'd have my fields in the a column and "as prev_" in column B and my fields again in column c. In column d i'd have a column.

Then use concatanate in column e and merge them together, making sure to include spaces. Then cut and paste this into your sql code. I've also used this method to make case statements for the same field and other longer codes i need to do for each field in a multihundred field table.

This creates the list of fields with a given prefix

select
    name + ' as prefix.' + name + ','
from sys.columns where object_id = object_id('mytable')
order by column_id

I use to_jsonb function in PostgreSQL 13 to get all fields in joined table as a single column.

select
  TABLE_A.*,
  to_jsonb(TABLE_B.*) as b,
  to_jsonb(TABLE_C.*) as c
from TABLE_A
left join TABLE_B on TABLE_B.a_id=TABLE_A.id
left join TABLE_C on TABLE_C.a_id=TABLE_A.id
where TABLE_A.id=1

As a result you will get the number of TABLE_A columns plus b and c columns:

You just need to parse the b and c columns to convert them to an object.

Same response as the very good 'PHP (Wordpress) function' but coded for CakePHP 4.3. Place in src/Controller/Component/MyUtilsComponent.php

<?php

namespace App\Controller\Component;

use Cake\Controller\Component;
use Cake\Datasource\ConnectionManager;

class MyUtilsComponent extends Component
{
    public static function prefixedTableFieldsWildcard(string $table, string $alias, string $connexion = 'default'): string
    {
        $c = ConnectionManager::get($connexion);
        $columns = $c->execute("SHOW COLUMNS FROM $table");
        $field_names = [];
        foreach ($columns as $column) {
            $field_names[] = $column['Field'];
        }

        $prefixed = [];
        foreach ($field_names as $field_name) {
            $prefixed[] = "`{$alias}`.`{$field_name}` AS `{$alias}.{$field_name}`";
        }
        return implode(', ', $prefixed);
    }
}

Tests and usage

    function testPrefixedTableFieldsWildcard(): void
    {
        $fields = MyUtilsComponent::prefixedTableFieldsWildcard('metas', 'u', 'test');
        $this->assertEquals('`u`.`id` AS `u.id`, `u`.`meta_key` AS `u.meta_key`, `u`.`meta_value` AS `u.meta_value`, `u`.`meta_default` AS `u.meta_default`, `u`.`meta_desc` AS `u.meta_desc`', $fields,);
    }

You would think that over 13 years Microsoft would have put this in. It would be incredibly useful for debugging purposes. What i've gotten in the habit of doing is selecting the columns I think I want to compare and then putting an * at the end to catch anything else I might want to look at.

select a.breed, a.size, p.breed, p.size,a.,p. from animal a join pet p on a.breed=p.breed

anyway you get the idea.