API Documentation

Context / Registry

class libres.context.registry.Registry[source]

Holds a number of contexts, managing their creation and defining the currently active context.

A global registry instance is found in libres:

from libres import registry

Though if global state is something you need to avoid, you can create your own version of the registry:

from libres.context.registry import create_default_registry
registry = create_default_registry()
register_context(name, replace=False)[source]

Registers a new context with the given name and returns it.


Creates the default registry for libres.

class libres.context.core.Context(name, registry=None, parent=None, locked=False)[source]

Used throughout Libres, the context holds settings like the database connection string and services like the json dumps/loads functions that should be used.

Contexts allow consumers of the Libres library to override these settings / services as they wish. It also makes sure that multiple consumers of Libres can co-exist in a single process, as each consumer must operate on it’s own context.

Libres holds all contexts in libres.registry and provides a master_context. When a consumer registers its own context, all lookups happen on the custom context. If that context can provide a service or a setting, it is used.

If the custom context can’t provide a service or a setting, the master_context is used instead. In other words, the custom context inherits from the master context.

Note that contexts not meant to be changed often. Classes talking to the database usually cache data form the context freely. That means basically that after changing the context you should get a fresh Scheduler instance or call clear_cache().

A context may be registered as follows:

from libres import registry
my_context = registry.register_context('my_app')

See also Registry

__init__(name, registry=None, parent=None, locked=False)[source]
class libres.context.core.ContextServicesMixin[source]

Provides access methods to the context’s services. Expects the class that uses the mixin to provide self.context.

The results are cached for performance.


Clears the cache of the mixin.


Closes the current session.


Returns the current session.

class libres.context.core.StoppableService[source]

Services inheriting from this class have their stop_service method called when the service is discarded.

Note that this only happens when a service is replaced with a new one and not when libres is stopped (i.e. this is not a deconstructor).

class libres.context.session.SessionProvider(dsn, engine_config={}, session_config={})[source]

Global session utility. It provides a SERIALIZABLE session to libres. If you want to override this provider, be sure to set the isolation_level to SERIALIZABLE as well.

If you don’t do that, libres might run into errors as it assumes and tests against SERIALIZABLE connections!

__init__(dsn, engine_config={}, session_config={})[source]

Returns the postgres version in a tuple with the first value being the major version, the second being the minor version.

Uses it’s own connection to be independent from any session.


Called by the libres context when the session provider is being discarded (only in testing).

This makes sure that replacing the session provider on the context doesn’t leave behind any idle connections.

Database Access

class libres.db.scheduler.Scheduler(context, name, timezone)[source]

The Scheduler is responsible for talking to the backend of the given context to create reservations. It is the main part of the API.

__init__(context, name, timezone)[source]

Initializeds a new Scheduler instance.


The libres.context.core.Context this scheduler should operate on. Acquire a context by using libres.context.registry.Registry.register_context().


The name of the Scheduler. The context name and name of the scheduler are used to generate the resource uuid in the database. To access the data you generated with a scheduler use the same context name and scheduler name together.


A single scheduler always operates on the same timezone. This is used to determine what a whole day means for example (given that a whole day starts at 0:00 and ends at 23:59:59).

Dates passed to the scheduler that are not timezone-aware are assumed to be of this timezone!

This timezone cannot change after allocations have been created! If it does, a migration has to be written (as of yet no such migration exists).

allocate(dates, partly_available=False, raster=5, whole_day=False, quota=None, quota_limit=0, grouped=False, data=None, approve_manually=False)[source]

Allocates a spot in the sedate.

An allocation defines a timerange which can be reserved. No reservations can exist outside of existing allocations. In fact any reserved slot will link to an allocation.


The datetimes to allocate. This can be a tuple with start datetime and an end datetime object, or a list of tuples with start and end datetime objects.

If the datetime objects are timezone naive they are assumed to be of the same timezone as the scheduler itself.


If an allocation is partly available, parts of its daterange may be reserved. So if the allocation lats from 01:00 to 03:00, a reservation may be made from 01:00 to 02:00.

if partly_available if False, it may only be reserved as a whole (so from 01:00 to 03:00 in the aforementioned example).

If partly_available is True, a raster may be specified. See raster.


If an allocation is partly available a raster defines the granularity with which a reservation can be made.

For example: a raster of 15min will ensure that reservations are at least 15 minutes long and start either at :00, :15, :30 or :45).

By default, we use a raster of 5, which means that reservations may not be shorter than 5 minutes and will snap to 00:05, 00:10, 00:15 and so on.

For performance reasons it is not possible to create reservations shorter than 5 minutes. If you need that, this library is not for you.


If true, the hours/minutes of the given dates are ignored and they are made to span a whole day (relative to the scheduler’s timezone).


The number of times this allocation may be ‘over-reserved’. Say you have a concert and you are selling 20 tickets. The concert is on saturday night, so there’s only one start and end date. But there are 20 reservations/tickets that can be made on that allocation.

By default, an allocation has a quota of one and may therefore only be reserved once.


The number of times a reservation may ‘over-reserve’ this allocation. If you are selling tickets for a concert and set the quota_limit to 2, then you are saying that each customer may only acquire 2 tickets at once.

If the quota_limit is 0, there is no limit, which is the default.


Creates a grouped allocation. A grouped allocation is an allocation spanning multiple date-ranges that may only be reserved as a whole.

An example for this is a college class which is scheduled to be given every tuesday afternoon. A student may either reserve a spot for the class as a whole (including all tuesday afternoons), or not at all.

If the allocation has only one start and one end date, the grouped parameter has no effect.

If allocate is called with multiple dates, without grouping, then every created allocation is completely independent.

By default, allocations are not grouped.


A dictionary of your own chosing that will be attached to the allocation. Use this for your own data. Note that the dictionary needs to be json serializable.

For more information see Custom JSON Serializer/Deserializer.


If true, reservations must be approved before they generate reserved slots. This allows for a kind fo waitinglist/queue that forms around an allocation, giving an admin the possiblity to pick the reservations he or she approves of.

If false, reservations trigger a reserved slots immediatly, which results in a first-come-first-serve kind of thing.

Manual approval is a bit of an anachronism in Libres which might be removed in the future. We strongly encourage you to not use this feature and to just keep the default (which is False).

allocations_by_reservation(token, id=None)[source]

Returns the allocations for the reservation if it was approved, pending reservations return nothing. If you need to get the allocation a pending reservation might be targeting, use _target_allocations in model.reservation.


This function approves an existing reservation and writes the reserved slots accordingly.

Returns a list with the reserved slots.

availability(start=None, end=None)[source]

Goes through all allocations and sums up the availability.

change_quota(master, new_quota)[source]

Changes the quota of a master allocation.

Fails if the quota is already exhausted.

When the quota is decreased a reorganization of the mirrors is triggered. Reorganizing means eliminating gaps in the chain of mirrors that emerge when reservations are removed:

Initial State: 1 (master) Free 2 (mirror) Free 3 (mirror) Free

Reservations are made: 1 (master) Reserved 2 (mirror) Reserved 3 (mirror) Reserved

A reservation is deleted: 1 (master) Reserved 2 (mirror) Free <– !! 3 (mirror) Reserved

Reorganization is performed: 1 (master) Reserved 2 (mirror) Reserved <– !! 3 (mirror) Free <– !!

The quota is decreased: 1 (master) Reserved 2 (mirror) Reserved

In other words, the reserved allocations are moved to the beginning, the free allocations moved at the end. This is done to ensure that the sequence of generated uuids for the mirrors always represent all possible keys.

Without the reorganization we would see the following after decreasing the quota:

The quota is decreased: 1 (master) Reserved 3 (mirror) Reserved

This would make it impossible to calculate the mirror keys. Instead the existing keys would have to queried from the database.

change_reservation(token, id, new_start, new_end, quota=None)[source]

Allows to change the timespan of a reservation under certain conditions:

  • The new timespan must be reservable inside the existing allocation. (So you cannot use this method to reserve another allocation)
  • The referenced allocation must not be in a group.

Returns True if a change was made.

Just like revoke_reservation, this function raises an event which includes a send_email flag and a reason which may be used to inform the user of the changes to his reservation.

change_reservation_time(token, id, new_start, new_end)[source]

Kept for backwards compatibility, use change_reservation() instead.


Returns the reservations that fullfill the restrictions imposed by change_reservation_time.

Pass a list of reservation tokens to further limit the results.


Clones the scheduler. The result will be a new scheduler using the same context, name, settings and attributes.


Denies a pending reservation, removing it from the records and sending an email to the reservee.


WARNING: Completely removes any trace of the records managed by this scheduler. That means all reservations, reserved slots and allocations!

free_allocations_count(master_allocation, start, end)[source]

Returns the number of free allocations between master_allocation and it’s mirrors.


The allocations managed by this scheduler / resource.


The reservations managed by this scheduler / resource.


The reserved_slots managed by this scheduler / resource.


Returns True if any of the allocations require manual approval.

remove_reservation(token, id=None)[source]

Removes all reserved slots of the given reservation token.

Note that removing a reservation does not let the reservee know that his reservation has been removed.

If you want to let the reservee know what happened, use revoke_reservation.

The id is optional. If given, only the reservation with the given token AND id is removed.

remove_unused_allocations(start, end)[source]

Removes all allocations without reservations between start and end and returns the number of allocations that were deleted.

Groups which are partially inside the daterange are not included.

reordered_keylist(allocations, new_quota)[source]

Creates the map for the keylist reorganzation.

Each key of the returned dictionary is a resource uuid pointing to the resource uuid it should be moved to. If the allocation should not be moved they key-value is None.

reservation_targets(start, end)[source]

Returns a list of allocations that are free within start and end. These allocations may come from the master or any of the mirrors.

reserve(email, dates=None, group=None, data=None, session_id=None, quota=1, single_token_per_session=False)[source]

Reserves one or many allocations. Returns a token that needs to be passed to approve_reservations() to complete the reservation.

That is to say, Libres uses a two-step reservation process. The first step is reserving what is either an open spot or a place on the waiting list (see approve_manually of allocate()).

The second step is to actually write out the reserved slots, which is done by approving an existing reservation.

Most checks are done in the reserve functions. The approval step only fails if there’s no open spot.

This function returns a reservation token which can be used to approve the reservation in approve_reservation.

Usually you want to just short-circuit those two steps:


Each reservation must be associated with an email. That is, a user.


The dates to reserve. May either be a tuple of start/end datetimes or a list of such tuples.


The allocation group to reserve. dates``and ``group are mutually exclusive.


A dictionary of your own chosing that will be attached to the reservation. Use this for your own data. Note that the dictionary needs to be json serializable.

For more information see Custom JSON Serializer/Deserializer.


An uuid that connects the reservation to a browser session.

Together with libres.db.queries.Queries.confirm_reservations_for_session() this can be used to create a reservation shopping card.

By default the session_id is None, meaning that no browser session is associated with the reservation.


The number of allocations that should be reserved at once. See quota in allocate().


If True, all reservations of the same session shared the same token, though that token will differ from the session id itself.

This only applies if the reserve function is called multiple times with the same session id. In this case, subsequent reserve calls will re-use whatever token they can find in the table.

If there’s no existing reservations, a new token will be created. That also applies if a reservation is created, deleted and then another is created. Because the last reserve call won’t find any reservations it will create a new token.

So the shared token is always the last token returned by the reserve function.

Note that this only works reliably if you set this parameter to true for all your reserve calls that use a session.

reserved_slots_by_reservation(token, id=None)[source]

Returns all reserved slots of the given reservation. The id is optional and may be used only return the slots from a specific reservation matching token and id.


The resource that belongs to this scheduler. The resource is a uuid created from the name and context of this scheduler, based on the namespace uuid defined in settings.uuid_namespace

search_allocations(start, end, days=None, minspots=0, available_only=False, whole_day='any', groups='any', strict=False)[source]

Search allocations using a number of options. The date is split into date/time. All allocations between start and end date within the given time (on each day) are included.

For example, start=01.01.2012 12:00 end=31.01.2012 14:00 will include all allocations in January 2012 which OVERLAP the given times. So an allocation starting at 11:00 and ending at 12:00 will be included!

WARNING allocations not matching the start/end date may be included if they belong to a group from which a member is included!

If that behavior is not wanted set ‘strict’ to True or set ‘include_groups’ to ‘no’ (though you won’t get any groups then).

Allocations which are included in this way will return True in the following expression:

getattr(allocation, ‘is_extra_result’, False)


Include allocations starting on or after this date.


Include allocations ending on or before this date.


List of days which should be considered, a subset of: ([‘mo’, ‘tu’, ‘we’, ‘th’, ‘fr’, ‘sa’, ‘su’])

If left out, all days are included.


Minimum number of spots reservable.


If True, unavailable allocations are left out (0% availability). Default is False.


May have one of the following values: ‘yes’, ‘no’, ‘any’

If yes, only whole_day allocations are returned. If no, whole_day allocations are filtered out. If any (default), all allocations are included.

Any is the same as leaving the option out.


‘any’ if all allocations should be included. ‘yes’ if only group-allocations should be included. ‘no’ if no group-allocations should be included.

See allocation.in_group to see what constitutes a group


Set to True if you don’t want groups included as a whole if a groupmember is found. See comment above.


Creates the tables and indices required for libres. This needs to be called once per database. Multiple invocations won’t hurt but they are unnecessary.

class libres.db.queries.Queries(context)[source]

Contains helper methods independent of the resource (as owned by scheduler.Scheduler)

Some contained methods require the current context (for the session). Some contained methods do not require any context, they are marked as staticmethods.

static allocations_in_range(query, start, end)[source]

Takes an allocation query and limits it to the allocations overlapping with start and end.

static availability_by_allocations(allocations)[source]

Takes any iterator with alloctions and calculates the availability. Counts missing mirrors as 100% free and returns a value between 0-100 in any case. For single allocations check the allocation.availability property.

availability_by_day(start, end, resources)[source]

Availability by range with a twist. Instead of returning a grand total, a dictionary is returned with each day in the range as key and a tuple of availability and the resources counted for that day.

WARNING, this function should run as linearly as possible as a lot of records might be processed.

availability_by_range(start, end, resources)[source]

Returns the availability for the given resources in the given range. The exposure is used to check if the allocation is visible.

confirm_reservations_for_session(session_id, token=None)[source]

Confirms all reservations of the given session id. Optionally confirms only the reservations with the given token. All if None.


Goes through all reservations and returns the session ids of the unconfirmed ones which are older than the given expiration date. By default the expiration date is now - 15 minutes.

Note that this method goes through ALL RESERVATIONS OF THE CURRENT SESSION. This is NOT limited to a specific context or scheduler.


Removes all reservations which have an expired session id. By default the expiration date is now - 15 minutes.

See find_expired_reservation_sessions()

Note that this method goes through ALL RESERVATIONS OF THE CURRENT SESSION. This is NOT limited to a specific context or scheduler.

remove_reservation_from_session(session_id, token)[source]

Removes the reservation with the given session_id and token.


Events are called by the libres.db.scheduler.Scheduler whenever something interesting occurs.

The implementation is very simple:

To add an event:

from libres.modules import events

def on_allocations_added(context_name, allocations):


To remove the same event:


Events are called in the order they were added.

class libres.modules.events.Event[source]

Event subscription. By http://stackoverflow.com/a/2022629

A list of callable objects. Calling an instance of this will cause a call to each item in the list in ascending order by index.

libres.modules.events.on_allocations_added = []

Called when an allocation is added, with the following arguments:

Context:The libres.context.core.Context used when adding the allocations.
Allocations:The list of libres.db.models.Allocation allocations to be commited.
libres.modules.events.on_reservation_time_changed = []

Called when a reservation’s time changes , with the following arguments:

Context:The libres.context.core.Context used when changing the reservation time.
Reservation:The libres.db.models.Reservation reservation whose time is changing.
Old_time:A tuple of datetimes containing the old start and the old end.
New_time:A tuple of datetimes containing the new start and the new end.
libres.modules.events.on_reservations_approved = []

Called when a reservation is approved, with the following arguments:

Context:The libres.context.core.Context used when approving the reservation.
Reservations:The list of libres.db.models.Reservation reservations being approved.
libres.modules.events.on_reservations_confirmed = []

Called when a reservation bound to a browser session is confirmed, with the following arguments:

context:The libres.context.core.Context used when confirming the reservation.
reservations:The list of libres.db.models.Reservation reservations being confirmed.
session_id:The session id that is being confirmed.
libres.modules.events.on_reservations_denied = []

Called when a reservation is denied, with the following arguments:

Context:The libres.context.core.Context used when denying the reservation.
Reservations:The list of libres.db.models.Reservation reservations being denied.
libres.modules.events.on_reservations_made = []

Called when a reservation is made, with the following arguments:

Context:The libres.context.core.Context used when adding the reservation.
Reservations:The list of libres.db.models.Reservation reservations to be commited. This is a list because one reservation can result in multiple reservation records. All those records will have the same reservation token and the same reservee email address.
libres.modules.events.on_reservations_removed = []

Called when a reservation is removed, with the following arguments:

Context:The libres.context.core.Context used when removing the reservation.
Reservations:The list of libres.db.models.Reservation reservations being removed.


class libres.db.models.Allocation(**kwargs)[source]

Describes a timespan within which one or many timeslots can be reserved.

There’s an important concept to understand before working with allocations. The resource uuid of an alloction is not always pointing to the actual resource.

A resource may in fact be a real resource, or an imaginary resource with a uuid derived from the real resource. This is a somewhat historical artifact.

If you need to know which allocations belong to a real resource, the mirror_of field is what’s relevant. The originally created allocation with the real_resource is also called the master-allocation and it is the one allocation with mirror_of and resource being equal.

When in doubt look at the managed_* functions of the scheduler.Scheduler class.


A simple constructor that allows initialization from kwargs.

Sets attributes on the constructed instance using the names and values in kwargs.

Only keys that are present as attributes of the instance’s class are allowed. These could be, for example, any mapped columns or relationships.

align_dates(start=None, end=None)[source]

Aligns the given dates to the start and end date of the allocation.

all_slots(start=None, end=None)[source]

Returns the slots which exist with this timespan. Reserved or free.


True if reservations for this allocation must be approved manually.


Returns the availability in percent.


Partitions the space between start and end into blocks of either free or reserved time. Each block has a percentage representing the space the block occupies compared to the size of the whole allocation.

The blocks are ordered from start to end. Each block is an item with two values. The first being the percentage, the second being true if the block is reserved.

So given an allocation that goes from 8 to 9 and a reservation that goes from 8:15 until 8:30 we get the following blocks:

    (25%, False),
    (25%, True),
    (50%, False)

This is useful to divide an allocation block into different divs on the frontend, indicating to the user which parts of an allocation are reserved.

contains(start, end)[source]

Returns true if the the allocation contains the given dates.


Creates a new copy of this allocation.

count_slots(start=None, end=None)[source]

Returns the number of slots which exist with this timespan. Reserved or free.


Custom data reserved for the user


Returns the end plus one microsecond in either the timezone given or the timezone on the allocation.


Returns the start in either the timezone given or the timezone on the allocation.


The end of this allocation. Must be timezone aware. This date is rastered by the allocation’s raster. The end date is stored with an offset of minues one microsecond to avoid overlaps with other allocations. That is to say an allocation that ends at 15:00 really ends at 14:59:59.999999

find_spot(start, end)[source]

Returns the first free allocation spot amongst the master and the mirrors. Honors the quota set on the master and will only try the master if the quota is set to 1.

If no spot can be found, None is returned.

free_slots(start=None, end=None)[source]

Returns the slots which are not yet reserved.


Group uuid to which this allocation belongs to. Every allocation has a group but some allcations may be the only one in their group.


the id of the allocation, autoincremented


True if the event is in any group.

is_available(start=None, end=None)[source]

Returns true if the given daterange is completely available.


True if the allocation is a master allocation.


True if available separately (as opposed to available only as part of a group).


True if the allocation does not exist in the database, and is not about to be written to the database. If an allocation is transient it means that the given instance only exists in memory.

See: http://www.sqlalchemy.org/docs/orm/session.html #quickie-intro-to-object-states http://stackoverflow.com/questions/3885601/ sqlalchemy-get-object-instance-state

limit_timespan(start, end, timezone=None)[source]

Takes the given timespan and moves the start/end date to the closest reservable slot. So if 10:00 - 11:00 is requested it will

  • on a partly available allocation return 10:00 - 11:00 if the raster allows for that
  • on a non-partly available allocation return the start/end date of the allocation itself.

The resulting times are combined with the allocations start/end date to form a datetime. (time in, datetime out -> maybe not the best idea)


resource of which this allocation is a mirror. If the mirror_of attribute equals the resource, this is a real resource see models.Allocation for more information

overlaps(start, end)[source]

Returns true if the allocation overlaps with the given dates.


Partly available allocations may be reserved partially. How They may be partitioned is defined by the allocation’s raster.


Returns the pending reservations query for this allocation. As the pending reservations target the group and not a specific allocation this function returns the same value for masters and mirrors.


Number of times this allocation may be reserved


Maximum number of times this allocation may be reserved with one single reservation.


the resource uuid of the allocation, may not be an actual resource see models.Allocation for more information


Returns the master/mirrors group this allocation is part of.

If ‘imaginary’ is true, inexistant mirrors are created on the fly. those mirrors are transient (see self.is_transient)


The start of this allocation. Must be timezone aware. This date is rastered by the allocation’s raster.


The timezone this allocation resides in.


True if the allocation is a whole-day allocation.

A whole-day allocation is not really special. It’s just an allocation which starts at 0:00 and ends at 24:00 (or 23:59:59‘999). Relative to its timezone.

As such it can actually also span multiple days, only hours and minutes count.

The use of this is to display allocations spanning days differently.

class libres.db.models.ReservedSlot(**kwargs)[source]

Describes a reserved slot within an allocated timespan.


A simple constructor that allows initialization from kwargs.

Sets attributes on the constructed instance using the names and values in kwargs.

Only keys that are present as attributes of the instance’s class are allowed. These could be, for example, any mapped columns or relationships.

class libres.db.models.Reservation(**kwargs)[source]

Describes a pending or approved reservation.


A simple constructor that allows initialization from kwargs.

Sets attributes on the constructed instance using the names and values in kwargs.

Only keys that are present as attributes of the instance’s class are allowed. These could be, for example, any mapped columns or relationships.


Returns the end plus one microsecond (nicer display).


Does nothing but to form a nice pair to display_end.


Returns the timespans targeted by this reservation.

The result is a list of Timespan timespans. The start and end are the start and end dates of the referenced allocations.

The timespans are ordered by start.

class libres.db.models.reservation.Timespan(start, end)



default: None

The data source name to connect to the right database. For example: postgresql+psycopg2://user:password@localhost:5432/database


default: UUID(‘49326ef9-fbc0-4ac0-9508-b0bbd75d42f7’)

The namespace used by the scheduler to create uuids out of the context and the name. You usually don’t want to change this. You really do not want to change it once you have created records using the scheduler - otherwise you will lose the connection between your context/name and the specific record in the database.

Just leave it really.