Introduction to tables

What is a table?

A table displays some data in a tabular way, using cells that are arranged in rows and columns.

In a Lino application you describe tables using Python classes. These Python classes are a general description of how to lay out your data and can be used for different front ends. The same table description is used to render data interactively as a grid panel or on a printable document as a table.

Don't mix up models and tables: while your models describe how data is to be stored in the database, your tables describe how data is to be presented to users.

Lino's "tables" are roughly equivalent of Django's "views". With Lino you don't need to write views because Lino writes them for you. Actually a Lino table corresponds only to one type of Django's views, sometimes referred to as "tabular" or "list" views. The other class of views are "detail" views, for which you are going to define Layouts (we'll talk about these later).

In Lino we differentiate between database tables and virtual tables. Database tables get their data directly from the database using a Django model. Virtual tables have no model, they get their data programmatically.

Implementation note: database tables are subclasses of lino.core.dbtables.Table (generally imported via its shortcut dd.Table), virtual tables are subclasses of lino.core.tables.VirtualTable (generally imported via its shortcut dd.VirtualTable). The classes have a common abstract base class lino.core.tables.AbstractTable.

The remainder of this tutorial concentrates on database tables, virtual tables have a tutorial on their own.

Illustration

To illustrate this, we will have a look at the lino_book.projects.tables demo application.

Here are the database models:

from __future__ import unicode_literals
from lino.api import dd
from django.db import models
from django.core.exceptions import ValidationError


@dd.python_2_unicode_compatible
class Author(dd.Model):
    first_name = models.CharField("First name", max_length=50)
    last_name = models.CharField("Last name", max_length=50)
    country = models.CharField("Country", max_length=50, blank=True)
    
    def __str__(self):
        return "%s, %s" % (self.last_name, self.first_name)


class Book(dd.Model):
    author = dd.ForeignKey(Author, blank=True, null=True)
    title = models.CharField("Title", max_length=200)
    published = models.IntegerField(
        "Published",
        help_text="The year of publication")
    price = models.DecimalField("Price", decimal_places=2, max_digits=10)


    def full_clean(self):
        super(Book, self).full_clean()
        if self.published > 2000 and self.price < 5:
            price = dd.format_currency(self.price)
            msg = "A book from {} for only {}!".format(
                self.published, price)
            raise ValidationError(msg)
                

        

And here are the tables:

from lino.api import dd


class Authors(dd.Table):
    model = 'Author'
    column_names = 'first_name last_name country'

    detail_layout = """
    first_name last_name country
    BooksByAuthor
    """


class Books(dd.Table):
    model = 'Book'
    column_names = 'author title published *'
    hide_sums = True


class RecentBooks(Books):
    column_names = 'published title author'
    order_by = ['published']


class BooksByAuthor(Books):
    master_key = 'author'
    column_names = 'published title'
    order_by = ['published']

Tables can be defined either together with the database models in your models.py, or in a separate file named desktop.py.

Tables are subclasses of dd.Table. You don't need to instantiate them, Lino loads them automatically at startup and they are globally available at runtime in the lino.api.rt module.

>>> from lino import startup
>>> startup('lino_book.projects.tables.settings')
>>> from lino.api import rt, dd
>>> rt.models.tables.Books
lino_book.projects.tables.desktop.Books
>>> issubclass(rt.models.tables.Books, dd.Table)
True

There can be more than one table for a given model, but each table has exactly one model as its data source. That model is specified in the model attribute of the table. For every database model there should be at least one table. Lino will generate a default table for models that have no table at all.

Much information about your table is automatically extracted from the model: the columns correspond to the fields of your database model. The header of every column is the verbose_name of its field. The values in a column are of same data type for each row. So Lino knows all these things from your models.

The rows of a table can be sorted and filtered. These are things which are done in Django on a QuerySet. Lino doesn't reinvent the wheel here and just forwards them to their corresponding Django methods: order_by, filter and exclude.

But here is something you cannot express on a Django model: which columns are to be shown, and how they are ordered. This is defined by the column_names attribute, a simple string with a space-separated list of field names.

Tables can hold information which goes beyond a model or a queryset. For example we set hide_sums to True on the Books table because otherwise Lino would display a sum for the "published" column.

Designing your tables

Database models are usually named in singular form, tables in plural form.

Tables may inherit from other tables (e.g. BooksByAuthor inherits from Books: it is basically a list of books, with the difference that it shows only the books of a given author.

The recommended place for defining tables is in a separate file desktop.py. You might define your tables together with the models in your models.py file, but in that case your application has no chance to support responsive design.

As a rule of thumb you can say that you need one table for every grid view used in your application. Each table is a subclass of dd.Table.

To define tables, you simply need to declare their classes. Lino discovers and analyzes them when it initializes. Tables never get instantiated.

Each table class must have at least one class attribute model which points to the model on which this table will "work". Every row of a table represents an instance of its model. (This is true only for database tables. Lino also has virtual tables, we will talk about them in a later tutorial.

Since tables are normal Python classes they can use inheritance. In our code BooksByAuthor inherits from Books. That's why we don't need to explicitly specify a model attribute for BooksByAuthor.

BooksByAuthor is an example of a slave table. It shows the books of a given Author. This given Author is called the "master" of these Books. We also say that a slave table depends on its master.

Lino manages this dependency almost automatically. The application developer just needs to specify a class attribute master_key. This attribute, when set, must be a string containing the name of a ForeignKey field which must exist in the Table's model.

A table can defined attributes like filter and order_by which you know from Django's QuerySet API.

A table is like a grid widget, it has attributes like column_names which describe how to display it to the user.

But the table is even more than the definition of a grid widget. It also has attributes like detail_layout which tells it how to display the detail of a single record in a form view.

Try also to work through the API docs, knowing that lino.core.dbtables.Table inherits from lino.core.tables.AbstractTable who inherits from lino.core.actors.Actor.

Using tables without a web server

An important thing with tables is that they are independent of any front end. You define them once, and you can use them on the console, in a script, in a testcase, in a web interface or in a GUI window.

At this point of our tutorial, we won't yet fire up a web browser (because we want to explain a few more concepts like menus and layouts before we can do that), but we can already play with our data using Django's console shell:

$ python manage.py shell

The first thing you do in a shell session is to import everything from lino.api.shell:

>>> from lino.api.shell import *

This imports especially a name rt which points to the lino.api.rt module. rt stands for "run time" and it exposes Lino's runtime API. In our first session we are going to use the show method and the actors object.

>>> rt.show(tables.Authors)
... 
============ =========== =========
 First name   Last name   Country
------------ ----------- ---------
 Douglas      Adams       UK
 Albert       Camus       FR
 Hannes       Huttner     DE
============ =========== =========

So here is, our Authors table, in a testable console format!

And here is the Books table:

>>> rt.show(tables.Books)
... 
================= ====================================== ===========
 author            Title                                  Published
----------------- -------------------------------------- -----------
 Adams, Douglas    Last chance to see...                  1990
 Adams, Douglas    The Hitchhiker's Guide to the Galaxy   1978
 Huttner, Hannes   Das Blaue vom Himmel                   1975
 Camus, Albert     L'etranger                             1957
================= ====================================== ===========

These were so-called master tables. We can also show the content of slave tables :

>>> adams = tables.Author.objects.get(last_name="Adams")
>>> rt.show(tables.BooksByAuthor, adams)
... 
=========== ======================================
 Published   Title
----------- --------------------------------------
 1978        The Hitchhiker's Guide to the Galaxy
 1990        Last chance to see...
=========== ======================================

Before going on, please note that the preceding code snippets are tested as part of Lino's test suite. This means that as a core developer you can run a command (inv test in case you are curious) which will parse the source file of this page, execute every line that starts with >>> and verifies that the output is the same as in this document. If a single dot changes, the test "fails" and the developer will find out the reason.

Writing test cases is an important part of software development. It might look less funny than developing cool widgets, but actually these are part of analyzing and describing how your users want their data to be structured. Which is the more important part of software development.

Defining a web interface

The last piece of the user interface is the menu definition, located in the __init__.py file ot this tutorial:

from lino.api import ad, _


class Plugin(ad.Plugin):
    verbose_name = _("Tables")

    def setup_main_menu(self, site, profile, m):
        m = m.add_menu(self.app_label, self.verbose_name)
        m.add_action('tables.Authors')
        m.add_action('tables.Books')

    

Every plugin of a Lino application can define its own subclass of lino.core.plugin.Plugin, and Lino instantiates these objects automatically a startup, even before importing your database models.

You might ask "Why can't we just define the menu commands in our settings.py or the models.py files? That question goes beyond the scope of this tutorial

Note that a plugin corresponds to what Django calls an application. More about this in Introduction to plugins.

About this tutorial

You can interactively play around with the little application used in this tutorial:

$ go tables
$ python manage.py runserver

Some screenshots:

../../_images/12.png ../../_images/21.png

The fixtures/demo.py file contains the data we used to fill our database:

from lino.api.shell import *
from lino.utils.instantiator import Instantiator


def objects():
    author = Instantiator(
        'tables.Author', 'first_name last_name country').build
    adams = author("Douglas", "Adams", "UK")
    yield adams
    camus = author("Albert", "Camus", "FR")
    yield camus
    huttner = author("Hannes", "Huttner", "DE")
    yield huttner

    book = Instantiator('tables.Book', 'title author published price').build
    yield book("Last chance to see...", adams, 1990, '9.90')
    yield book("The Hitchhiker's Guide to the Galaxy", adams, 1978, '19.90')
    yield book("Das Blaue vom Himmel", huttner, 1975, '14.90')
    yield book("L'etranger", camus, 1957, '6.90')
    # yield book("Book", camus, 2001, '4.90')

Exercises

Explore the application and try to extend it: change things in the code and see what happens.