Skip to content
This repository


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
branch: master
Fetching contributors…


Cannot retrieve contributors at this time

file 398 lines (292 sloc) 13.773 kb
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397

This describes, somewhat pedantically and in a non-helpful order,
parts of what makes up a PyWebApp. Known-open-questions are marked
with "FIXME".

Process Initialization

This is the process that the container must go through when
starting/activating an application.

You should instantiate ``app = pywebapp.PyWebApp()``

Then you should create the settings module with
``app.setup_settings()``. This creates a module ``websettings``.

Next you must activate all the services. This is something your tool
itself must do. At its most basic you do::

    websettings.add_setting(service_name, settings)

For example::

    websettings.add_setting('mysql', {'username': 'root', ...})

This creates a value ``websettings.mysql`` (but the function does
other checks for validity). There are some helpers you can use (in
````), but these are a convenience to you.

**After** you have setup the services you call ``app.activate_path()``
to set ``sys.path``. This also may import code that uses settings, so
it is important to setup services first.

Get the app from ``app.wsgi_app``

Process Configuration

The ``websettings`` module contains configuration from the
host/container of the application that is being sent to the
application. Anything can go in here, including ad hoc settings, but
some settings are expected.

    The path to the configuration (a directory). May be None. Must
    be set.

    The "base" hostname. The application may be on a wildcard domain
    or something like that, but this is at least one hostname that
    will point back to the application. It may be the only hostname.
    This like the Host header: ``domain:port``.

    The expected scheme, generally either ``http`` or ``https``.
    ``environ['wsgi.url_scheme']`` also of course must be set.
    Generally this is set to https when the container is forcing all
    requests to be https.

    A path where log files can be written. It is simply writable, and
    is entirely under control of the application. (FIXME: maybe a
    couple names here could be reserved by the container? Or... maybe
    best not?)

Also some environmental context should be set properly:

current directory:
    Must be the application root

    A temporary directory. This should be application-private, and as
    such a suitable place to put cache files and the like. The
    container should try to clear this out at appropriate times (like
    an update).

Also expect that over time specific settings will be documented and
validated by this module. E.g., the contents of ``websettings.mysql``
may be standardized specifically.

FIXME: some "version" of this specifically itself should go in here,
and also be possible to require or request in the application

Application Description

Configuration files are generally YAML throughout the system. The
application description is in the root of the application directory,
and is called ``app.yaml``.

    This is the basic platform of the application. ``wsgi`` is the
    only value we've thought through. Something like Tornado requires
    the server itself to be run, an application entry point doesn't
    work; it would be another kind of platform (e.g.,

    A Python file. It does not need to be importable, and may have
    for example ``-`` in its name, or use an extension like ``.wsgi``
    instead of ``.py``.

    This file when exec'd should produce a global variable
    ``application``. This is a WSGI application.

     A single path or a list of paths that should be added to
    ``sys.path``. All paths will get priority over system-wide paths,
    and ``.pth`` files will be interpreted. Also if a
    ```` file exists in any path it will be exec'd.
    Multiple ```` files may be exec'd! By default
    ``lib/pythonX.Y``, ``lib/pythonX.Y/site-packages``, and
    ``lib/python`` will be loaded. It is best to just use the last

    A path that will contain static files. These will be available
    *at the root* of the application, so if you do ``static_path:
    static/`` then ``static/favicon.ico`` will be available at the URL
    path ``/favicon.ico``. If you want a file at, for instance,
    ``/static/blank.gif`` you must put it in

    This value defaults to ``static/``

    FIXME: there should be a way to set mimetypes

    Indicates the Python versions supported. (FIXME: just
    Setuptools-style requirement, e.g., >=2.7,<3.0 ?)

    This is the hosting platform you require (a list of options).
    Generally ``posix`` and ``win`` are the options.

    These are settings specific to Debian and Ubuntu systems.

    Packages that should be installed on Debian/Ubuntu systems.

    Values specific to RPM-like systems (Redhat, CentOS, etc).

    Packages that should be installed on RPM-based systems. (FIXME:
    often specific packages are needed, and there isn't a central
    repository). (FIXME: maybe we should allow ``rpm.requirements``
    being a ``requirements.txt`` file containing anything that isn't
    available in a package, but less than the global ``requirements``

    A path to a pip ``requirements.txt`` file. (FIXME: does deb/rpm
    configuration takes precedence over this? Or do we just make sure
    those packages are installed first, so that any already-met
    requirements don't then need to be reinstalled?)

    This relates to configuration that the application requires. This
    is for applications that are not self-configured. Configuration
    is simply a directory, which may contain one or more files, in any
    format, to be determined by the application itself.

    Note that configuration should be things that a deployer might
    want to change in some useful fashion. E.g., a blog title.

    If true then configuration is required.

    This is a template for creating a valid configuration (which the
    deployer may then edit). What kind of template is not yet

    Probably it would contain some kind of structured description of
    what parameters the template requires, and then a routine that
    given the parameters will create a directory structure.

    This checks a configuration for validity (presumably after the
    deployer edits it). This is a command. Not fully defined.

    This is a relative path which points to the default configuration
    if one is not given. If this value is provided then
    ``config.required`` won't really matter, as the application will
    always have at least its default configuration.

    This contains a number of named services. It is up to the
    container to interpret these and setup the services. (FIXME:
    clearly this needs to be expanded.)


Several configuration values are "commands", that is: something that
can be executed. All commands are run in the activated environment
(i.e., after ``sys.path`` has been updated, and with service

Commands take one of several formats:

    This is a URL that will be fetched. It may be fetched through an
    artificial WSGI request (i.e., not over-the-wire HTTP).

    A URL starts with ``url:`` or simply anything that starts with
    ``/``. E.g., ``/__heartbeat__`` indicates a request to that URL.
    FIXME: also there should be a way to tell that it is being called
    as a command, not as an external URL (e.g., special environ key).

Python script:
    This is a script that will be run. It will be run with
    ``execfile()``, and in it ``__name__ == '__main__'``, so any
    normal script can be used.

    A Python script starts with ``pyscript:`` or any path that *does
    not* start with ``/`` and ends in ``.py``. All paths of course
    are relative to the root of the application.

General script:
    This is a script that is run in a subprocess, and could be for
    instance a shell script. FIXME: we would need to define the

    The script will be run with the current working directory of the
    application root. It will be run with something akin to
    ``os.system()``, i.e., as a shell script.

    General scripts must start with ``script:``.

Python functions:
    This is a function to be called. The function cannot take any

    A Python function must start with ``pyfunc:`` and have either a
    complete dotted-notation path, or ``module:object.attr``
    (Setuptools-style). Also anything that does not start with ``/``
    and is contains only valid Python identifiers and ``.`` will
    automatically be considered a Python function.

Commands can only generally return success or failure, plus readable
messages. The failure case is specific:

* For URLs, 2xx is success, all other status is a failure. Output is
  the body of the response.

* For a Python script, calling ``SystemExit`` (or ``sys.exit()``) with
  a non-zero code is failure, an exception is failure, otherwise it is
  success. The output is what is printed to ``sys.stdout`` and

* For a General script, exit code and stdout/stderr.

* For a Python function, an exception is a failure, all else is
  success. Output is stdout/stderr or a string return value (if both,
  the string is appended to output). (FIXME: non-string, truish
  return value?)

The command environment for scripts should be:

  This environmental variable points to the application root.

current directory:
  This also should(?) be at the application root.

In the case of exceptions, the output value is preserved and the
``str()`` of the exception is added. (FIXME: also the traceback?)

It may be sensible to allow a combination of a rich object (e.g.,
response with headers, list of interleaved stdout/stderr, etc) and a
string fallback.

FIXME: General scripts and URLs can implicitly have arguments (URLs
having the query string), but the others can't (maybe Python scripts
also can have arguments?). Maybe we should allow shell-quoted
arguments to all commands (except URLs).


At different stages of an application's deployment lifecycle

    Called when an application is first installed. This would be the
    place to create database tables, for instance.

    Called before an update is applied. This is called in the context
    of the previous installation/version.

    Called when an application is updated, called after the update in
    the context of the new version.

    Called before deleting an application.

    Called to check if an application is alive; must be low in
    resource usage. A URL is most preferable for this parameter.

    Called to check if an application is in good shape. May do
    integrity checks on data, for instance. May be high in resource

    Called to check if the configuration is valid.

    Can be called to confirm that the environment is properly
    configured, for instance to check that all necessary command-line
    programs are available. This check is optional, the environment
    need not run it.

If only one of ``install`` or ``update`` are defined, then they are
used interchangeably. E.g., if ``install`` is defined, it is called
on updates. Or if only ``update`` is defined, it is called on install.

Kinds of Services

Services are things the provider provides, and can represent a variety
of things. All application state must be represented through
services. As such you can't do much of interest without at least some


This represents just a place to keep files. The files don't do
anything, they are to be read and written by the application.

The configuration looks like::

    websettings.files = {'dir': <directory>}

You can write to this directory. An optional key ``"quota"`` is the
most data you are allowed to put in this directory (in bytes).


This is a place to keep files that you want served up. These files
take precedence over your own application!

The directory layout starts with the domain of the request, or
``default`` for any/all domains. So if you want to write out a file
that will be served up in ``/user-content/public.html`` then write it
to ``<public_files>/default/user-content/public.html``

Note that this has obvious security concerns, so you should write
things carefully.

Configuration looks like::

    websettings.public_files = {"dir": <directory>}

``"quota"`` is also supported.


Several databases act similarly.

The configuration parameters that are generally necessary are kept in
a dictionary with these keys:

  The host that the database is on (e.g., ``"localhost"``)

  The port of the database

  The name of the database (e.g., ``db_1234``)

  The user to connect as. May be None.

  The password to connect with. May be None.

  Entirely optional, this is a second user that could be created for
  use during runtime (as opposed to application setup). This user
  might not have permission to create or delete tables, for instance.

  Accompanying password.
Something went wrong with that request. Please try again.