Ultipa filter judges whether a metadata meets the requirement by applying an expression against its property or schema. Features of Ultipa filter are:

• It is an expression encapsulated in curly braces { }, and yields a value of bool or null
• It usually (but not necessarily) calls the schema and/or property of current metadata
• It validates a metadata only if the expression returns true

When the expression yields neither bool nor null, convert the result to bool based on below principles:

Scenario A: General Filtering

In any query command, filter current metadata by calling constant and/or alias defined in previous statement. For example:

find().edges({time >= "2021-09-01 09:00:00"}) 
...

... as maxAge
find().nodes({age == maxAge}) 
...

The properties filtered in this scenario, e.g. the edge property time and node property age in the above code, can be directly used for queries without any acceleration processing.

Scenario B: Inter-Step Filtering

In template query, filter current metadata by calling prev_n、prev_e and/or alias defined earlier in currernt statement. For example:

n(as start).e()[3].n({level == start.level})
...

n().e({@transfer.time > prev_e.time})[3].n()
...

The properties filtered in this scenario, e.g. the node property level and edge property time in the above code, must be LTE-ed before being used for queries.

Parameters path_ascend() and path_descend() used in some path query commands serve a inter-step comparison purpose similar to the inter-step filtering of template query. They also require the subject property to be LTE-ed, just that the subject property is not used in the form of filter.

Scenario C: Full-Text Filtering

In any query command, filter the full-text index of current metadata using a key-word constant. For example:

find().nodes({~content CONTAINS "graph computing parallel"}) 
...

The full-text index filtered in this scenario, e.g. the content in the above code, must be created using command create().node_fulltext() before being used for queries.