<divclass="dev_page_bread_crumbs"><ulclass="breadcrumb clearfix"><li><ahref="/api">API</a></li><iclass="icon icon-breadcrumb-divider"></i><li><ahref="/api/offsets">Pagination in the API</a></li></ul></div>
<h1id="dev_page_title">Pagination in the API</h1>
<divid="dev_page_content"><!-- scroll_nav -->
<p>Lots of Telegram API methods provide access to potentially large lists of objects, which requires pagination.</p>
<p>In order to fetch only relevant subset of results for each request there is a number of available input parameters. Here is a list in order how they are applied in API.</p>
<p>A limit on the number of objects to be returned, typically between 1 and 100. When 0 is provided the limit will often default to an intermediate value like ~20.</p>
<p>For a few methods with mostly static data this parameter allows to skip <code>offset</code> elements from the beginning of list; negative values are allowed in some cases.</p>
<p>For most methods where results are real-time data (e.g. any chat history) <code>offset</code> value is not passed directly. Instead it is calculated from the passed <code>offset_id</code> and <code>add_offset</code> parameter values as <code>offsetFromID(offset_id) + add_offset</code>, where <code>offsetFromID(offset_id)</code> is a number of results from the beginning of list up to the result with ID <code>offset_id</code>, inclusive.</p>
<p>To further reduce the result subset, there is a mechanism to avoid fetching data if the resulting list hasn't changed from the one stored on client, similar to <ahref="https://en.wikipedia.org/wiki/HTTP_ETag">ETag</a>.</p>
<p>When the client has cached results for API request, it can calculate the <code>hash</code> value for it by taking the result IDs (message IDs or other fields with name <code>id</code>, or some extra fields in some cases) and using them to compute a 64-bit hash with the following algorithm:</p>
<p>The <code>>></code> operator is the unsigned right shift operator. </p>
<p>Note: in some cases, the <code>ids</code> array passed to the algorithm must contain strings (i.e. the shortcut name in business shortcuts, and so on...), in which case they must be transformed to longs by taking the first 8 bytes of the MD5 hash of the string (<strong>not</strong> in hex form) and treating it as a big-endian 64-bit long. </p>
<p>In some cases, if the result container already has a <code>hash</code> field, that can be used instead. </p>
<p>When the client passes a correct value, the API will return one of <code>*NotModified</code> constructors, e.g. <ahref="/constructor/messages.messagesNotModified">messages.messagesNotModified</a> instead of the actual results.</p>
<li><ahref="/method/messages.getHistory">messages.getHistory</a> supports all result navigation parameters including message ID hashes and except filters</li>
<li><ahref="/method/channels.getParticipants">channels.getParticipants</a> supports simple navigation using <strong>limit</strong> and <strong>offset</strong>, along with filtering and <code>hash</code> reducing using the user IDs of returned participants</li>
