Introduction to data views

What is a data view?

data view

A Python class that describes how to retrieve and display a set of data rows.

Each data view provides a set of display modes, a list of actions and layouts.

table view

Old word for a tabular data view.

tabular data view

A data view in a display mode that uses columns and rows to display the data (which is either grid mode or plain mode).

model-based data view

A data view that is connected to database model.

In a Lino application you describe data views using Python classes. These Python classes are an abstract description of how to lay out your data. Depending on the front end this leads to similar but not identical results.

A same data view is used to render data interactively as a grid panel or on a printable document as a table.

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

Lino's "data views" are more general than Django's "views". With Lino you don't need to write Django views because Lino writes them for you. Every data view usually causes multiple Django views.

In Lino we differentiate between model-based data views and virtual tables. The former get their data directly from the database using a Django model. The latter have no database model, they get their data programmatically.

Implementation note: model-based table views 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 model-based table views, 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 lino.api import dd
from django.db import models
from django.core.exceptions import ValidationError



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)


from .ui import *

And here are the data views:

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']

Data views must exist in your models.py, in the same namespace as your database models. But you might prefer to actually define them in a separate file and import them into your models.py by saying:

from .ui import *

We recommend to name such a file ui.py.

Note that using a desktop.py file is deprecated.

Table views are subclasses of dd.Table. You don't need to instantiate them, Lino discovers 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.ui.Books
>>> issubclass(rt.models.tables.Books, dd.Table)
True

There can be more than one data view for a given database model, but each data view has exactly one model as its data source. That model is specified in the model attribute. For every database model there should be at least one data view, otherwise Lino will generate a default data view for it.

Much information about your data view 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 things are done in Django on a QuerySet. Lino 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.

Data views can hold information that goes beyond a database 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 data

Database models are usually named in singular form, data views in plural form.

Date views may inherit from other data views (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.

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

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

Each data view must have at least one class attribute model, which points to the model on which this view 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 data views 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 of the data view's model.

A data view can define attributes like filter and order_by, which you know from Django's QuerySet API.

The columns of a data view

An important attribute of a data view is column_names, which describes the columns to show in

class lino.core.tables.AbstractTable

column_names = '*'

A string that describes the list of columns of this table.

Default value is '*', which means to show all columns.

Lino will automatically create a lino.core.layouts.ColumnsLayout from this. This string must not contain any newline characters because a ColumnsLayout's main panel descriptor must be horizontal.

See also setup_column() and get_column_names().

hidden_columns = frozenset()

If given, this is specifies the data elements that should be hidden by default when rendering this table. Example:

hidden_columns = "long_name expected_date"

Value : Application code should specify this as a single string containing a space-separated list of field names. Lino will automatically resolve this during server startup using lino.core.utils.fields_list(). The runtime value of this attribute is a set of strings, each one the name of a data element. Defaults to an empty set.

Inheritance : Note that this can be specified either on a Model or on a Table. Lino will make a union of both.

Wildcard columns

The asterisk ('*') in a column specifier is a wildcard and means "insert at this point all columns that have not been named explicitly". It can be combined with explicitly specified names. These wildcard columns

If '*' is not present in the string, only explicitly named columns will be available.

For example:

column_names = "name owner * date"

specifies that name and owner come first, followed by inserted columns and finally by date.

Virtual fields are not included as wildcard field unless they have lino.core.fields.VirtualField.wildcard_field set to True. This rule is for performance reasons. Some virtual fields a rather heavy (e.g. the navigation_panel must query the whole database to get all primary keys), and even when they are hidden, Lino has to include wildcard fields in the result because the end user might have enabled them.

Other table view attributes

But the table is even more than the description 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.

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 of 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.

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

Exercises

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

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

$ go tables
$ python manage.py runserver

Some screenshots:

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')

Glossary

wildcard column

A data element that has been inserted by a * and which is hidden by default. See Wildcard columns.

wildcard field

A database field that is candidate to becoming a wildcard column.