Parameters, Value

CASE is a user-defined function which maps every row of data in a data stream to a new value, and all the new values have the same structure type.

If data do not come from homologous columns: when used in WITH, case will first make a Cartesian Product Combination of all columns and its homologous columns before mapping; when used in RETURN, case will trim all columns and their homologous columns before mapping.

Syntax:

• Format：
  case
    when <condition1> then <projection1>
    when <condition2> then <projection2>
    ...
    else <other>
  end
  
• Parameters:（see table below）
• Value：ATTR、ARRAY、TABLE

The mapped value of data in every row with different types of <projection> and <other>:

Start to judge every data in a row from the first <condition>, output the mapped value once judge condition is satisfied and stop judge the conditions following.

• Abbreviated format: if all <conditions> are used to judge if a certain unified expression<uFormat> equals with other expressions <format1>, <format2>,... ，then the function can be put in this way:
  case <uFormat>
    when <format1> then <projection1>
    when <format2> then <projection2>
    ...
    else <other>
  end
  

Constant

Example: divide @customer nodes into 4 age groups based on birth years: 1980s, 1990s, and 2000s, count the number of people in each age group

find().nodes({@customer}) as nodes
with CASE
when year(nodes.birthday) <=> [1980, 1989] then "1980s"
when year(nodes.birthday) <=> [1990, 1999] then "1990s"
when year(nodes.birthday) <=> [2000, 2009] then "2000s"
else "others" END as ageGroup
group by ageGroup
return ageGroup, count(ageGroup)

Example: abbreviate the example above

find().nodes({@customer}) as nodes
with CASE floor(year(nodes.birthday)/10)
when 198 then "1980s"
when 199 then "1990s"
when 200 then "2000s"
else "others" END as ageGroup
group by ageGroup
return ageGroup, count(ageGroup)

Function

Example: calculate the actual monthly paydays in 2022, with 15th each month as planned payday and payday will be put off to the nearest Monday if it is a non-workday.

uncollect ["2022-1-15 0:0:0","2022-2-15 0:0:0","2022-3-15 0:0:0","2022-4-15 0:0:0","2022-5-15 0:0:0","2022-6-15 0:0:0","2022-7-15 0:0:0","2022-8-15 0:0:0","2022-9-15 0:0:0","2022-10-15 0:0:0","2022-11-15 0:0:0","2022-12-15 0:0:0"] as plan
return CASE day_of_week(plan)
when 1 then date_add(plan, 1, "day")
when 7 then date_add(plan, 2, "day")
else date_add(plan, 0, "day")
END

Analysis: the first condition in case adds 1 day to a planned payday that happens to be a Sunday, the second condition adds 2 days to a planned payday on Saturday , ELSE keeps the planned payday as it is by adding 0 day. Note: ELSE can not be followed by plan, for plan's structure type is string, date_add() can only be filled with time, so to keep data structure type consistent plan should be transformed to a time format.

Alias

Example: find one 1-step path from each card CA001, CA002, and CA005, count each path as [<transferCardID>, <amount>,<receivercardID>]; if a path do not exist, count it as [<CardID>, 0, "none"]

find().nodes({_id in ["CA001","CA002","CA005"]}) as cards
optional n(cards).le({@transfer} as e).n({@card} as n).limit(1)
return CASE n
when 0 then [cards._id, 0, "none"]
else [cards._id, e.amount, n._id] END

Analysis: 1) when using optional, if one of the template query does not return result, then conduct a psyeudo-path 0-0-0; 2）use 'CASE' to find the psyeudo-path 0-0-0 and map it as [<CardID>, 0, "none"], map the rest of paths as a table [<TransferorCardID>,<amount>,<ReceiverCardID>], note that data type should be consistent in each column.