Querying Views

View Object

class couchbase.views.iterator.View
__init__(parent, design, view, row_processor=None, streaming=0, include_docs=False, query=None, **params)

Construct a iterable which can be used to iterate over view query results.

  • parent (Connection) – The parent Connection object
  • design (string) – The design document
  • view (string) – The name of the view within the design document
  • row_processor (callable) – See row_processor for more details.
  • include_docs (boolean) –

    If set, the document itself will be retrieved for each row in the result. The default algorithm uses get_multi() for each page (i.e. every streaming results).

    The reduce family of attributes must not be active, as results fro reduce views do not have corresponding doc IDs (as these are aggregation functions).

  • streaming (bool) – Whether a streaming chunked request should be used. This is helpful for handling the view results in small chunks rather than loading the entire resultset into memory at once. By default, a single request is made and the response is decoded at once. With streaming enabled, rows are decoded incrementally.
  • query – If set, is a Query object. It is illegal to use this in conjunction with additional params
  • params – Extra view options. This may be used to pass view arguments (as defined in Query) without explicitly constructing a Query object. It is illegal to use this together with the query argument. If you wish to ‘inline’ additional arguments to the provided query object, use the query’s update() method instead.

This object is an iterator - it does not send out the request until the first item from the iterator is request. See __iter__() for more details on what this object returns.

Simple view query, with no extra options:

# c is the Connection object.

for result in View(c, "beer", "brewery_beers"):
    print("emitted key: {0}, doc_id: {1}"
            .format(result.key, result.docid))

Execute a view with extra query options:

# Implicitly creates a Query object

view = View(c, "beer", "by_location",

Pass a Query object:

q = Query(
        ["21st_ammendment_brewery_cafe", Query.STRING_RANGE_END]

view = View(c, "beer", "brewery_beer", query=q)

Add extra parameters to query object for single call:

view = View(c, "beer", "brewery_beer",
            query=q.update(debug=True, copy=True))

Include documents with query:

view = View(c, "beer", "brewery_beer",
            query=q, include_docs=True)

for result in view:
    print("Emitted key: {0}, Document: {1}".format(
        result.key, result.doc.value))

Returns a row for each query. The type of the row depends on the row_processor being used.

Raise :


If an error was encountered while processing the view, and the on_error attribute was not set to continue.

If continue was specified, a warning message is printed to the screen (via warnings.warn and operation continues). To inspect the error, examine errors

Raise :


If this object was already iterated over and the last result was already returned.



Errors returned from the view engine itself


Number of total rows indexed by the view. This is the number of results before any filters or limitations applied. This is only valid once the iteration has started


An object to handle a single page of the paginated results. This object should be an instance of a class conforming to the RowProcessor interface. By default, it is an instance of RowProcessor itself.


The actual couchbase.connection.HttpResult object. Note that this is only the last result returned. If using paginated views, the view comprises several such objects, and is cleared each time a new page is fetched.


Name of the design document being used


Name of the view being queired


Whether documents are fetched along with each row


How many actual rows were returned from the server.

This is incremented each time a new request is made. Note this may be different from the amount of rows yielded by iterator from RowProcessor.handle_rows() if a custom row_processor is being used

Row Processing

class couchbase.views.iterator.RowProcessor
handle_rows(rows, connection, include_docs)

Preprocesses a page of rows.

  • rows (list) – A list of rows. Each row is a JSON object containing the decoded JSON of the view as returned from the server
  • connection – The connection object (pass to the View constructor)
  • include_docs – Whether to include documents in the return value. This is True or False depending on what was passed to the View constructor

an iterable. When the iterable is exhausted, this method will be called again with a new ‘page’.

class couchbase.views.iterator.ViewRow

This is the default class returned by the RowProcessor


The key emitted by the view’s map function (first argument to emit)


The value emitted by the view’s map function (second argument to emit). If the view was queried with reduce enabled, then this contains the reduced value after being processed by the reduce function.


This is the document ID for the row. This is always None if reduce was specified. Otherwise it may be passed to one of the get or set method to retrieve or otherwise access the underlying document. Note that if include_docs was specified, the doc already contains the document


If include_docs was specified, contains the actual couchbase.connection.Result object for the document.

Query Object

class couchbase.views.params.Query
__init__(passthrough=False, unrecognized_ok=False, **params)

Create a new Query object.

A Query object is used as a container for the various view options. It can be used as a standalone object to encode queries but is typically passed as the query value to View.

  • passthrough (boolean) – Whether passthrough mode is enabled
  • unrecognized_ok (boolean) – Whether unrecognized options are acceptable. See Circumventing Parameter Constraints.
  • params – Key-value pairs for view options. See view_options for a list of acceptable options and their values.
Raise :

couchbase.exceptions.ArgumentError if a view option or a combination of view options were deemed invalid.

update(copy=False, **params)

Chained assignment operator.

This may be used to quickly assign extra parameters to the Query object.


q = Query(reduce=True, full_sec=True)

# Someplace later

v = View(design, view, query=q.update(mapkey_range=["foo"]))

Its primary use is to easily modify the query object (in-place).

  • copy (boolean) – If set to true, the original object is copied before new attributes are added to it
  • params – Extra arguments. These must be valid query options.

A Query object. If copy was set to true, this will be a new instance, otherwise it is the same instance on which this method was called


Returns an encoded form of the query

This document explains the various view options, and how they are treated by the Couchbase library.

Many of the view options correspond to those listed here http://www.couchbase.com/docs/couchbase-manual-2.0/couchbase-views-querying-rest-api.html

Note that these explain the view options and their values as they are passed along to the server.

These attributes are available as properties (with get and set) and can also be used as keys within a constructor.

Result Range and Sorting Properties

The following properties allow you to

  • Define a range to limit your results (i.e. between foo and bar)
  • Define a specific subset of keys for which results should be yielded
  • Reverse the sort order
class couchbase.views.params.Query

Specify the range based on the contents of the keys emitted by the view’s map function.

Server Option:Maps to both startkey and endkey
Value Type:Range Value of JSON Value elements

The result output depends on the type of keys and ranges used.

One may specify a “full” range (that is, an exact match of the first and/or last key to use), or a partial range where the start and end ranges specify a subset of the key to be used as the start and end. In such a case, the results begin with the first key which matches the partial start key, and ends with the first key that matches the partial end key.

Additionally, keys may be compound keys, i.e. complex data types such as lists.

You may use the STRING_RANGE_END to specify a wildcard for an end range.

Match all keys that start with “a” through keys starting with “f”:

q.mapkey_range = ["a", "f"+q.STRING_RANGE_END]
q.inclusive_end = True

If you have a view function that looks something like this:

function(doc, meta) {
    if (doc.city && doc.event) {
        emit(doc.city, doc.event)

Then you may query for all events in a specific city by using:

q.mapkey_range = [["Reno"], ["Reno", q.STRING_RANGE_END]]

The empty element r[0][1] means to match the lowest string value for each result. The special last element r[1][1] is a constant which expands to the highest acceptable unicode value. This means to match any string in the second element as long as it is below this value (which effectively means any string).

As such, the results may look like:

ViewRow(key=["Reno", "Air Races"], ...),
ViewRow(key=["Reno", "Street Vibrations"], ...)

# etc.
STRING_RANGE_END = u'\u0fff'

Highest acceptable unicode value

Server Option:Maps to both startkey_docid and endkey_docid
Value Type:Range Value of String elements.

Specify the range based on the contents of the keys as they are stored by set(). These are returned as the “Document IDs” in each view result.

You must use this attribute in conjunction with mapkey_range option. Additionally, this option only has any effect if you are emitting duplicate keys for different document IDs. An example of this follows:


c.set("id_1", { "type" : "dummy" })
c.set("id_2", { "type" : "dummy" })
# ...
c.set("id_9", { "type" : "dummy" })


// This will emit "dummy" for ids 1..9

function map(doc, meta) {

Only get information about "dummy" docs for IDs 3 through 6:

q = Query()
q.mapkey_range = ["dummy", "dummy" + Query.STRING_RANGE_END]
q.dockey_range = ["id_3", "id_6"]
q.inclusive_end = True


Apparently, only the first element of this parameter has any effect. Currently the example above will start returning rows from id_3 (as expected), but does not stop after reaching id_6.

Server Option:key
Value Type:JSON Value

Limit the view results to those keys which match the value to this option exactly.


function(doc, meta) {
    if (doc.type == "brewery") {
    } else {
        emit([doc.brewery_id, meta.id]);


q.mapkey_single = ["abbaye_de_maredsous"]

Note that as the map function can return more than one result with the same key, you may still get more than one result back.

Server Option:keys
Value Type:JSON Array

Like mapkey_single, but specify a sequence of keys. Only rows whose emitted keys match any of the keys specified here will be returned.


q.mapkey_multi = [
    ["abbaye_de_maresdous", "abbaye_de_maresdous-8"],
    ["abbaye_do_maresdous", "abbaye_de_maresdous-10"]
Server Option:inclusive_end
Value Type:Boolean Type

Declare that the range parameters’ (for e.g. mapkey_range and dockey_range) end key should also be returned for rows that match it. By default, the resultset is terminated once the first key matching the end range is found.

Server Option:descending
Value Type:Boolean Type

Reduce Function Parameters

These options are valid only for views which have a reduce function, and for which the reduce value is enabled

class couchbase.views.params.Query
Server Option:reduce
Value Type:Boolean Type

Note that if the view specified in the query (to e.g. couchbase.connection.Connection.query()) does not have a reduce function specified, an exception will be thrown once the query begins.

Server Option:group
Value Type:Boolean Type

Specify this option to have the results contain a breakdown of the reduce function based on keys produced by map. By default, only a single row is returned indicating the aggregate value from all the reduce invocations.

Specifying this option will show a breakdown of the aggregate reduce value based on keys. Each unique key in the result set will have its own value.

Setting this property will also set reduce to True

Server Option:group_level
Value Type:Numeric Type

This is analoguous to group, except that it places a constraint on how many elements of the compound key produced by map should be displayed in the summary. For example if this parameter is set to 1 then the results are returned for each unique first element in the mapped keys.

Setting this property will also set reduce to True

Pagination and Sampling

These options limit or paginate through the results

class couchbase.views.params.Query
Server Option:skip
Value Type:Numeric Type


Consider using mapkey_range instead. Using this property with high values is typically inefficient.

Server Option:limit
Value Type:Numeric Type

Set an absolute limit on how many rows should be returned in this query. The number of rows returned will always be less or equal to this number.

Control Options

These do not particularly affect the actual query behavior, but may control some other behavior which may indirectly impact performance or indexing operations.

class couchbase.views.params.Query
Server Option:stale

Specify the (re)-indexing behavior for the view itself. Views return results based on indexes - which are not updated for each query by default. Updating the index for each query would cause much performance issues. However it is sometimes desirable to ensure consistency of data (as sometimes there may be a delay between recently-updated keys and the view index).

This option allows to specify indexing behavior. It accepts a string which can have one of the following values:

  • ok

    Stale indexes are allowable. This is the default. The constant STALE_OK may be used instead.

  • false

    Stale indexes are not allowable. Re-generate the index before returning the results. Note that if there are many results, this may take a considerable amount of time (on the order of several seconds, typically). The constant STALE_UPDATE_BEFORE may be used instead.

  • update_after

    Return stale indexes for this result (so that the query does not take a long time), but re-generated the index immediately after returning. The constant STALE_UPDATE_AFTER may be used instead.

A Boolean Type may be used as well, in which case True is converted to "ok", and False is converted to "false"

Server Option:on_error
Value Type:A string of either "stop" or "continue". You may use the symbolic constants ONERROR_STOP or ONERROR_CONTINUE

This parameter is a server-side option indicating how long a given node should wait for another node to respond. This does not directly set the client-side timeout.

Server Option:connection_timeout
Value Type:Numeric Type
Server Option:debug
Value Type:Boolean Type

If enabled, various debug output will be dumped in the resultset.

Server Option:full_set
Value Type:Boolean Type

If enabled, development views will operate over the entire data within the bucket (and not just a limited subset).

Value Type For Options

Different options accept different types, which shall be enumerated here

Boolean Type

Options which accept booleans may accept the following Python types:

  • Standard python bool types, like True and False
  • Numeric values which evaluate to booleans
  • Strings containing either "true" or "false"

Other options passed as booleans will raise an error, as it is assumed that perhaps it was passed accidentally due to a bug in the application.

Numeric Type

Options which accept numeric values accept the following Python types:

  • int, long and float objects
  • Strings which contain values convertible to said native numeric types

It is an error to pass a bool as a number, despite the fact that in Python, bool are actually a subclass of int.

JSON Value

Options which accept JSON values accept native Python types (and any user- defined classes) which can successfully be passed through json.dumps.

Do not pass an already-encoded JSON string, and do not URI-escape the string either - as this will be done by the option handling layer (but see Circumventing Parameter Constraints for a way to circumvent this)

Note that it is perfectly acceptable to pass JSON primitives (such as numbers, strings, and booleans).

JSON Array

Options which accept JSON array values should be pass a Python type which can be converted to a JSON array. This typically means any ordered Python sequence (such as list and tuple). Like JSON Value, the contents of the list should not be URI-escaped, as this will be done at the option handling layer


Options which accept strings accept so-called “semantic strings”, specifically; the following Python types are acceptable:

  • str and unicode objects
  • int and long objects

Note that bool, none and other objects are not accepted - this is to ensure that random objects passed don’t simply end up being repr()‘d and causing confusion in your view results.

If you have a custom object which has a __str__ method and would like to use it as a string, you must explicitly do so prior to passing it as an option.

Range Value

Range specifiers take a sequence (list or tuple) of one or two elements.

If the sequence contains two items, the first is taken to be the start of the range, and the second is taken to be its (non-inclusive) end

If the sequence contains only a single item, it is taken to be the start of the range, and no end will be specified.

To specify a range which has an end but not a start, pass a two-element sequence with the first element being an UNSPEC value.

The type of each element is parameter-specific.

Unspecified Value

Conventionally, it is common for APIs to treat the value None as being a default parameter of some sort. Unfortunately since view queries deal with JSON, and None maps to a JSON null, it is not possible for the view processing functions to ignore None.

As an alternative, a special constant is provided as UNSPEC. You may use this as a placeholder value for any option. When the view processing code encounters this value, it will discard the option-value pair.

Convenience Constants

These are convenience value constants for some of the options

params.ONERROR_CONTINUE = 'continue'
params.ONERROR_STOP = 'stop'
params.STALE_OK = 'ok'
params.STALE_UPDATE_BEFORE = 'false'
params.STALE_UPDATE_AFTER = 'update_after'
params.UNSPEC = <Placeholder>

Circumventing Parameter Constraints

Sometimes it may be necessary to circumvent existing constraints placed by the client library regarding view option validation.

For this, there are passthrough and allow_unrecognized options which may be set in order to allow the client to be more lax in its conversion process.

These options are present under various names in the various view query functions.

  • Passthrough

    Passthrough removes any conversion functions applied. It simply assumes values for all options are strings, and then encodes them rather simply

  • Allowing Unrecognized Options

    If a newer version of a server is released has added a new option, older versions of this library will not know about it, and will raise an error when it is being used. In this scenario, one can use the ‘allow unrecognized’ mode to add extra options, with their values being treated as simple strings.

    This has the benefit of providing normal behavior for known options.