Icon

SQL operations

Terms

Operation: An operation in a data connection. This takes the place of a "table" in SQL for relational databases.

Data source: part of a query that generates results. This is either an operation, a subquery, or a join statement.

Result set: the list of results, or rows, produced by a query. In Transposit, each result is a JSON object or array.

Column: a field in a result.

Select statement

Select queries return data from one or more data sources. In Transposit, nearly all queries will be select queries, regardless of what the underlying API HTTP method is.

Select statement syntax:

SELECT <* or column selection or JSON template>
FROM <operation or subquery or join>
WHERE <predicate>
EXPAND BY <columns>
ORDER BY <columns>
LIMIT <number>

The SELECT clause is required.

The clauses WHERE, EXPAND BY, ORDER BY and LIMIT can be used only if a FROM clause is used.

The clauses must appear in the order specified above.

Order of execution

  1. FROM
  2. WHERE
  3. EXPAND BY
  4. SELECT
  5. ORDER BY
  6. LIMIT

Select clause

The SELECT clause can be used for manipulating the structure and values of each item in the data source.

The SELECT clause gets as input a JSON object or JSON array from the data source and produces a JSON object or a JSON array.

The SELECT clause supports three ways to manipulate an item:

Select star

The * symbol selects the entire result without modifying it:

SELECT *
FROM connection.operation

Column selection

Column selection can be used to construct a JSON object with specific keys that will appear in the top level of the object.

Column selection cannot construct a nested JSON object or a JSON array; to construct these items use a JSON template.

The syntax for column selection is:

SELECT <column-expression> AS <column-alias>, <column-expression> AS <column-alias>, ...

<column-expression> describes how to construct a value in the output JSON object and can be one of the following:

  • <path>
  • <operation-alias>.<path>
  • <literal-value>
  • <binary-expression>

<path> describes the location of a value inside a JSON object or array. <path> contains one or more dot-separated keys/field names that describe the lookup chain of fields inside the input JSON object. The last item in <path> can be .*. Bracket notation with array indexes and string keys is also supported.

Examples of valid <path>s:

  • <key>
  • <key-1>.<key-2>. ... <key-N>
  • <key>.*
  • <key-1>.<key-2>. ... <key-N>.*
  • <key>[0]
  • <key-1>[0].<key-2>
  • <key-1>[0].<key-2>.*
  • <key-1>[0].<key-2>[1].<key-3>
  • [<key>]

<key> is an identifier.

<operation-alias> is an identifier.

<literal-value> is a number, string or boolean value.

<binary-expression> represents basic math operations (for numbers) or string concatenation (of strings) and can be one of the following:

  • <column-expression> + <column-expression>
  • <column-expression> - <column-expression>
  • <column-expression> * <column-expression>
  • <column-expression> / <column-expression>

AS <column-alias> is optional. <column-alias> is an identifier.

Result construction

Column selection constructs a JSON object for each item in the data source based on the specified <column-expression>s and <column-alias>s.

The <column-expression>s and <column-alias>s are used in the same order as they appear in the query.

Each <column-expression> will be calculated and resolved to a value. The resolved value can be a scalar value (number, string or boolean), JSON object or JSON array.

When resolving a <path>, lookups are done recursively into the JSON object for each path component. If no value is found at that <path>, the value is considered 'not found'.

A .* at the end of a <path> works as a 'spread' operator, copying each key at the current location into the result object.

For <operation-alias>.<path> the resolve process will resolve the specified <path> only under the results from the operation that was marked with the specified alias.

If no operation was marked with <operation-alias> the resolve will stop and the value will be considered as 'not found'.

For <binary-expression> the resolve process will evaluate the expression to produce a value.

If the types of the operands is not the same the resolve process will produce an error and the entire query will fail.

If the types of the operand is string, only the + operator is allowed, if a different operator is used the resolve process will produce an error and the entire query will fail.

If the resolve process produces a valid value, this value will be added to the output JSON object with a key, the key is selected in the following way:

  • If <column-alias> is specified it will be used as the key.
  • If <column-expression> is a <path> that does not end with .*, the last <key> will be used as the key.
  • If <column-expression> is a <path> that ends with .*, the keys and values under the last <key> will all be copied into the result object.
  • Otherwise the entire <column-expression> will be used as the key. It's recommended to use <column-alias> in this case.

If the same key is used more than once the last value will be used and will override any previous values that had the same key.

If the value is resolved to null or 'not found' this value will not be added it to the output JSON object.

Examples

Selecting a single value:

SELECT col1
FROM connection.operation

Will generate an array with one element: a JSON object with a single key:

[
{
"col1": ...
},
...
]

Selecting multiple values:

SELECT col1, col2, col3
FROM connection.operation

Will generate a JSON object with multiple keys:

[
{
"col1": ...,
"col2": ...,
"col3": ...
},
...
]

Selecting a nested value:

To access a value inside a nested object you can use a . (dot) or [<key>] (bracket notation) as a separator between the nested object keys.

For the following result:

[
{
"nested": {
"object": {
"value": "myValue"
}
}
}
]

The queries:

SELECT nested.object.value
FROM connection.operation
SELECT nested['object']['value']
FROM connection.operation

Will generate a JSON object with the key and value of the item at the specific path:

 [
{
"value": "myValue"
},
...
]

Array indexing is also supported. For example, for the following result:

[
{
"nested": ["value0", "value1", "value2"]
}
]

The query:

SELECT nested[0], nested[1]
FROM connection.operation

Will generate a JSON object of:

[
{
"0": "value0",
"1": "value1"
}
]

Changing the key:

You can use column alias to change the key of a value:

SELECT col1 AS foo
FROM connection.operation

Will generate a JSON object with the key foo, the value will be the value of col1:

[
{
"foo": ...
},
...
]

Using operation aliases:

When an operation (or subquery) is named with an alias, the alias can later be used as a qualifier at the beginning of the path to define exactly where to do the lookup for the path (the results of which operation or subquery to use). This is particularly useful in join queries, where the query has more than one data source.

SELECT T.col1
FROM connection.operation AS T

Will generate a JSON object with a single key:

[
{
"col1": ...
},
...
]

Literal value:

Numbers, strings and booleans can be used directly in a column selector. If no alias is provided, the literal value is used as both the key and value:

SELECT 7, 7 as value1, 'seven' as value2, true as value3

Will generate:

[
{
"7": 7,
"value1": 7,
"value2": "seven",
"value3": true
}
]

Binary expressions:

Binary expressions can be used for basic math operation (for numbers) or string concatenation (of strings). Binary expressions can use any combination of paths, literal values, and nested binary expressions. Parentheses can be used to define the order of operations.

The query:

SELECT (20 + 3) * 2 as value

Will generate:

[
{
"value": 46
}
]

With columns:

SELECT (col1 + 10) * nested.object.value1 as value
FROM connection.operation

Will use the values of col1 and nested.object.value1 to calculate the value of the expression.

Selecting all values in an object:

To access all key/value in a nested object you can use a .* in the end of the path that contains the values.

If the result has the format:

[
{
"nested": {
"object": {
"value1": ...,
"value2": ...,
"value3": ...
}
}
}
]

The query:

SELECT nested.object.*
FROM connection.operation

Will generate a JSON object with the all the keys and values in the specific path:

[
{
"value1": ...,
"value2": ...,
"value3": ...
},
...
]

JSON templates

JSON templates are a more generic way to construct JSON objects or arrays as the output item.

To construct a JSON object use the syntax:

SELECT { <key>: <json-value>, ... }

To construct a JSON array use the syntax:

SELECT [ <json-value>, ... ]

<key> is an identifier. For convenience, JSON templates do not require the use of string quotes around the key unless the <key> contains spaces, illegal characters or keywords.

<json-value> can be one of the following:

  • <path> - a column selection
  • <operation-alias>.<path> - a column selection with operation alias qualifier
  • <literal-value> - number, string or boolean
  • <binary-expression> - a calculated expression
  • <json-object> - construct a nested object
  • <json-array> - construct a nested array

<path>, <operation-alias>.<path>, <literal-value> and <binary-expression> are the same as in column selection. The only difference is that .* at the end of <path> is not allowed in JSON templates; use the spread operator instead.

<json-object> constructs a JSON object, the syntax is:

{ <key-1>: <json-value-1>, <key-2>: <json-value-2>, ... <key-N>: <json-value-N> }

<json-array> constructs a JSON array, the syntax is:

[ <json-value-1>,  <json-value-2>, ... <json-value-N>]

Spread operator

The spread operator expands a JSON object into a JSON object or a JSON array into a JSON array (similar to <path>.* in column selection).

Spread JSON object use:

SELECT { ... obj }

Spread JSON array use:

SELECT [ ... arr ]

The spread operator can be mixed with any other JSON template features.

Object construction

A JSON template constructs a JSON object or array for each item from the data source based on the specified template.

If the outer template is a <json-object> the item will be JSON object.

If the outer template is a <json-array> the item will be JSON array.

The values inside the outer template can be any type.

The <key>s and <json-value>s are used in the same order as they appear in the template.

Each <json-value> will be calculated and resolved. The resolved value can be a scalar value (number, string or boolean), JSON object or JSON array.

<path>, <operation-alias>.<path>, <immediate-value> and <binary-expression> are resolved the same as in column selection.

<json-object> and <json-array> are resolved recursively.

If the resolve process finds a valid value, this value will be added to the output JSON object or array. If the output item is a JSON object, the specified <key> will be used.

If the same key is used more than once, the last value will be used and will override any previous values that had the same key.

If the value is resolved to null or 'not found':

  • If the output item is a JSON object, this value will not be added it to the output item.
  • If the output item is a JSON array, this value will be added it to the output item as null.

Examples

Selecting a single value:

SELECT { col1: col1 }
FROM connection.operation

Will generate a JSON object with a single key:

[
{
"col1": ...
},
...
]

Selecting multiple values:

SELECT { col1: col1, col2: col2, col3: col3 }
FROM connection.operation

Will generate a JSON object with multiple keys:

[
{
"col1": ...,
"col2": ...,
"col3": ...
},
...
]

Selecting a nested value:

To access a value inside a nested object you can use a . (dot) as separator between the nested object keys.

For the following result:

[
{
"nested": {
"object": {
"value": "myValue"
}
}
}
]

The query:

SELECT { value: nested.object.value }
FROM connection.operation

Will generate a JSON object with the key and value of the item in the specific path:

[
{
"value": "myValue"
},
...
]

Changing the key:

The key is a part of the JSON template, so the same syntax can be used even if you want to use a different key:

SELECT { foo: col1 }
FROM connection.operation

Will generate a JSON object with the key foo, the value will be the value of col1:

[
{
"foo": ...
},
...
]

If the key contains spaces, illegal characters or is a keyword, it must be escaped:

SELECT { `key with spaces`: col1 }
FROM connection.operation

Will generate

[
{
"key with spaces": ...
},
...
]

Using an operation alias:

When an operation (or subquery) is named with an alias, the alias can later be used as a qualifier at the beginning of the path to define exactly where to do the lookup for the path (the results of which operation or subquery to use). This is particularly useful in join queries, where the query has more than one data source.

SELECT { col1: T.col1 }
FROM connection.operation AS T

Will generate a JSON object with a single key:

[
{
"col1": ...
},
...
]

Literal value:

Numbers, strings and booleans can be used as the value in a JSON template:

SELECT { value1: 7, value2: 'seven', value3: true }

Will generate:

[
{
"value1": 7,
"value2": "seven",
"value3": true
}
]

Binary expressions:

Binary expressions can be used for basic math operation (for numbers) or string concatenation (of strings). Binary expressions can use any combination of path, immediate values and nested binary expressions. Parentheses can be used to define the order of operations.

The query:

SELECT { value: (20 + 3) * 2 }

Will generate:

[
{
"value": 46
}
]

With columns:

SELECT { value: (col1 + 10) * nested.object.value1 }
FROM connection.operation

Will use the values of col1 and nested.object.value1 to calculate the value of the expression.

Constructing an array:

The previous examples showed how to construct an object as the output item. With JSON templates you can also construct an array as the output item.

SELECT [ col1 ]
FROM connection.operation

Will generate a JSON array with a single item (the item is the value of col1):

[
[
<VALUE OF col1>
],
...
]

All expressions shown in the previous examples can also be used: nested object, operation alias, immediate values, and binary expressions:

SELECT [ (col1 + 10) * nested.object.value1, T.col2, 7, 'seven', true ]
FROM connection.operation AS T

Constructing nested objects:

You can use JSON object and array templates recursively and construct any nested structure:

SELECT {
arr1: [ { col1: col1 }, { col2: col2 } ],
obj: {
bar: [
col3.nested.value1,
col3.nested.value2,
col3.nested.value3
]
}
}
FROM connection.operation

Selecting all values in an object or array:

To access all the values inside an object or array you can use the spread operator (...). Note that the inner type and the outer type must be the same: an object can be spread into a JSON object template and an array can be spread into a JSON array template:

If the results have the format:

[
{
"object": {
"value1": ...,
"value2": ...,
"value3": ...
},
"array": [ { "val1": ... }, { "val2": ... }, { "val3": ... } ]
}
]

The query:

SELECT { ... object }
FROM connection.operation

Will generate a JSON object with the all the keys and values under object:

[
{
"value1": ...,
"value2": ...,
"value3": ...
},
...
]

The query:

SELECT [ ... array ]
FROM connection.operation

Will generate a JSON array with the all the values under array:

[
[
{ "val1": ... },
{ "val2": ... },
{ "val3": ... }
],
...
]

From clause

The FROM clause of a query creates the result set that the other parts of the query will use. When the query is executed, the FROM clause runs first.

The FROM clause supports three types of data sources:

Operation

To get data from a single operation, you can use the operation directly in the FROM clause:

SELECT *
FROM connection.operation AS <operation-alias>

AS <operation-alias> is optional. <operation-alias> is an identifier.

Subquery

To manipulate the data set in multiple steps, you can use a subquery in the FROM clause:

SELECT *
FROM (SELECT * FROM connection.operation) AS <operation-alias>

AS <operation-alias> is optional. <operation-alias> is an identifier.

Both the outer and the inner queries can use any of the other clauses - FROM, WHERE, EXPAND BY, ORDER BY, LIMIT, SELECT.

Join

Joins can be used to merge the results of two or more operations that are related in some way.

SELECT *
FROM connection.operation_1 AS <operation-alias-1>
<join-type> JOIN connection.operation_2 AS <operation-alias-2>
ON <predicate>

<join-type> is optional and can be one of:

  • INNER
  • LEFT OUTER
  • RIGHT OUTER
  • FULL OUTER

If <join-type> is not specified the default is INNER.

In joins, AS <column-alias> is required. <operation-alias> is an identifier.

Where clause

The WHERE clause is used to specify input parameters to operations and filters on results. The syntax for the WHERE clause is:

WHERE <predicate>

A <predicate> describes the sequence of conditions that are applied to the data sources. Multiple <predicate>s can be recursively combined with boolean AND, OR, and NOT operators in the following format:

  • <condition>
  • <predicate> AND <predicate>
  • <predicate> OR <predicate>
  • NOT <predicate>

For instance:

SELECT * FROM <connection.operation>
WHERE <condition-1> AND NOT <condition-2> OR <condition-3> ...

Parentheses can also be used for grouping.

Each condition can be one of the following:

  • <name> <operator> <column-expression>
  • <name> <operator> <subquery>
  • <name> IN <literal-tuple>
  • <name-tuple> IN <nested-literal-tuple>

The <operator> can be one of the following:

  • = - equal to
  • != - not equal to
  • > - greater than
  • >= - greater than or equal to
  • < - less than
  • <= - less than or equal to
  • IN - equal to any in a list of values.

The <column-expression> is the same as what is used in column selection.

If a subquery is used, the <operator> must be = or IN. If the operator is =, the subquery must return a single result.

Input parameters and filters

A condition in a WHERE clause may either be passed as an input parameter or treated as a filter on the output of the data source. There is no difference in the syntax; Transposit automatically makes the determination based on the <name> of the condition: if it is defined as a parameter for the operation, the condition will be passed into the operation. Otherwise, the condition is treated as a filter and will remove rows from the result set that do not match the criteria.

When two input parameters are combined with an AND, both are passed to the underlying operation. However, if two input parameters are combined with an OR, it will result in two invocations of the operation. For instance:

SELECT * FROM <connection.operation> WHERE param1 = 'val1' AND param2 = 'val2'

Results in a single invocation of the operation that is passed 'val1' as param1 and 'val2' as param2.

SELECT * FROM <connection.operation> WHERE param1 = 'val1' OR param2 = 'val2'

Results in two invocations of the operation that are run in parallel. The first is passed 'val1' as param1 and is not passed a value for param2. The second invocation is passed 'val2' as param2 and is not passed a value for param1. When both invocations are complete, the result sets from each are concatenated together before other parts of the query are run.

Note that IN is equivalent to ORs of = operators, and thus may also result in multiple invocations of an operation. For instance:

SELECT * FROM <connection.operation> WHERE param1 IN ('val1', 'val2', 'val3')

Is equivalent to:

SELECT * FROM <connection.operation> WHERE param1 = 'val1' OR param1 = 'val2' OR param1 = 'val3'

Both will result in three parallel invocations of the operation.

Mapping and Tuples

The <literal-tuple> is one of:

  • (<val-1>, <val-2>, ...) - a list of literal values
  • ((<val-1-a>, <val-1-b>, ...), (<val-2-a>, <val-2-b>, ...), ...) - a list of nested tuples

The first can be used with the IN operator to map to a single parameter or result column:

WHERE col1 IN (<val-1>, <val-2>, ...)

The latter can be used to map to multiple parameters or result columns using a <name-tuple>:

WHERE (col1, col2) IN ((<col1-val1>, <col2-val1>), (<col1-val2>, <col2-val2>), ...)

The number of items in each nested tuple on the right side of the IN must match the number of items in the <name-tuple> on the left.

This sort of column mapping is also possible with subqueries.

WHERE (col1, col2, ..., colN) IN (SELECT col1, col2, ... colN FROM <connection.operation>)

The mapping is positional (first column is mapped to the first name in the tuple, and so on), so the names of the subquery columns do not have to match the names in the <name-tuple>. However, the number of columns in the subquery must match the number of items in the <name-tuple>.

Operation aliases

Like the <path> in column selection, the <name> in a WHERE condition can be prefixed with an operation alias to refer to the results of a named data source.

Expand by clause

The EXPAND BY clause expands (flattens) the items in a JSON array. This tends to be useful when working with APIs, where the relevant results may be nested inside one or more JSON objects.

The syntax for EXPAND BY is:

SELECT * FROM <connection.operation>
WHERE <predicate>
EXPAND BY <path-1> AS <column-alias>, <path-2> AS <column-alias>, ..., <path-N> AS <column-alias>

<column-path>

EXPAND BY works something like this:

For each result in the result set, for each <path> in the list of paths:

  1. Resolve the value at the path using the same mechanism as in column-selection.
  2. Check that this value is an array of items, otherwise skip this result.
  3. Create a new result for each item in the array. If there is no <column-alias>, the item is placed in the same position as the array it came from (replacing the array). Otherwise, the alias is used to place the item and the original array is left intact.
  4. Replace the top-level result set with this new list of results

Examples

Expand by a single column

If the results of connection.operation have the format:

[
{
"id": 1,
"vals": [1, 2]
},
{
"id": 2,
"vals": [3, 4]
}
]

The query:

SELECT * FROM connection.operation
EXPAND BY vals

Will create a row for each item in vals:

[
{ "id": 1, "vals": 1 },
{ "id": 1, "vals": 2 },
{ "id": 2, "vals": 3 },
{ "id": 2, "vals": 4 }
]

Expand by a single column with an alias

If the results of connection.operation have the format:

[
{
"id": 1,
"vals": [1, 2]
},
{
"id": 2,
"vals": [3, 4]
}
]

The query:

SELECT * FROM connection.operation
EXPAND BY vals as aliasedVals

Will create the following result set:

[
{
"id": 1,
"vals": [1, 2],
"aliasedVals": 1
},
{
"id": 1,
"vals": [1, 2],
"aliasedVals": 2
},
{
"id": 2,
"vals": [3, 4],
"aliasedVals": 3
},
{
"id": 2,
"vals": [3, 4],
"aliasedVals": 4
}
]

Notice that when an alias is used, the original array is left in the results.

Expand by a nested field

If the results of connection.operation have the format:

[
{
"id": 1,
"nested": {
"vals": [1, 2]
}
},
{
"id": 2,
"nested": {
"vals": [3, 4]
}
}
]

The query:

SELECT * FROM connection.operation
EXPAND BY nested.vals

Will create the following result set:

[
{
"id": 1,
"nested": {
"vals": 1
}
},
{
"id": 1,
"nested": {
"vals": 2
}
},
{
"id": 2,
"nested": {
"vals": 3
}
},
{
"id": 2,
"nested": {
"vals": 4
}
}
]

Notice that the nested structure is maintained.

Expand by multiple fields

If the results of connection.operation have the format:

[
{
"id": 1,
"letters": ["a", "b"],
"numbers": [1, 2]
},
{
"id": 2,
"letters": ["c", "d"],
"numbers": [3, 4]
}
]

The query:

SELECT * FROM connection.operation
EXPAND BY letters, numbers

Will create the following result set:

[
{
"id": 1,
"letters": "a",
"numbers": 1
},
{
"id": 1,
"letters": "a",
"numbers": 2
},
{
"id": 1,
"letters": "b",
"numbers": 1
},
{
"id": 1,
"letters": "b",
"numbers": 2
},
{
"id": 2,
"letters": "c",
"numbers": 3
},
{
"id": 2,
"letters": "c",
"numbers": 4
},
{
"id": 2,
"letters": "d",
"numbers": 3
},
{
"id": 2,
"letters": "d",
"numbers": 4
}
]

Expand by with missing fields

If the results of connection.operation have the format where vals is missing from one of the rows:

[
{
"id": 1,
"vals": [1, 2]
},
{
"id": 2,
},
{
"id": 3,
"vals": [3, 4]
}
]

The query:

SELECT * FROM connection.operation
EXPAND BY vals

Will create the following result set:

[
{
"id": 1,
"vals": 1
},
{
"id": 1,
"vals": 2
},
{
"id": 3,
"vals": 3
},
{
"id": 3,
"vals": 4
}
]

Notice that because the item with id: 2 did not have a field at the path vals, it does not show up in the result set.

Order by clause

The ORDER BY clause sorts the current set of rows.

The syntax for ORDER BY is:

SELECT * FROM <connection.operation>
WHERE <predicate>
ORDER BY <path-1> [ASC|DESC], <path-2> [ASC|DESC], ..., <path-N> [ASC|DESC]

<path> describes the location of a value inside a JSON object or array. <path> contains one or more dot-separated keys/field names that describe the lookup chain of fields inside the input JSON object. See Column selection for more details, or check the examples below.

At least one <path> is required.
The first path <path-1> will be the primary sort key, <path-2> will be the secondary sort key, etc.
The path must point to a scalar value in all the rows - number, string or boolean. All the values must have the same type (or null or undefined). Sorting on object or array values is not supported.

The sorting will be based on the natural order of the type with nulls/undefined first - numeric sorting for numeric values, lexicographic sorting for string values (based on the Unicode value of each character in the strings) and for boolean values - false first.

The ASC or DESC qualifiers are optional, the default will be ASC.
ASC - will sort the rows in ascending order.
DESC - will sort the rows in descending order.

ORDER BY is executed after SELECT - so any new field that was created in the SELECT clause can be used in the ORDER BY clause.

ORDER BY is executed before LIMIT - therefore in order to get the correct set of rows after the sorting and limit we must get all the data and only after that do the sorting and limit. In other words - if the operation is paginated, a query with ORDER BY will have to paginate through all the results and might increase the number of API calls. You can use subqueries to create a specific order of execution.

Examples

Order by a single column

If the results of connection.operation have the format:

[
{
"id": 1,
"name": "item 1"
},
{
"id": 2,
"name": "item 2"
}
]

The query:

SELECT * FROM connection.operation
ORDER BY id DESC

Will reorder the rows:

[
{ "id": 2, "name": "item 2" },
{ "id": 1, "name": "item 1" },
]

Order by multiple columns

If the results of connection.operation have the format:

[
{
"id": 1,
"type": "A"
},
{
"id": 2,
"type": "B"
},
{
"id": 3,
"type": "B"
},
{
"id": 4,
"type": "A"
}
]

The query:

SELECT * FROM connection.operation
ORDER BY type DESC, id ASC

Will reorder the rows:

[
[
{
"id": 2,
"type": "B"
},
{
"id": 3,
"type": "B"
},
{
"id": 1,
"type": "A"
},
{
"id": 4,
"type": "A"
}
]
]

Order by a nested value

If the results of connection.operation have the format:

[
{
"data": { values: [ 1, 2, 3 ] },
"name": "item 1"
},
{
"data": { values: [ 10, 20, 30 ] },
"name": "item 2"
},
{
"data": { values: [ 100, 200, 300 ] },
"name": "item 3"
}
]

The query:

SELECT * FROM connection.operation
ORDER BY data.values[1] DESC

Will reorder the rows:

[
{
"data": {
"values": [ 100, 200, 300 ]
},
"name": "item 3"
},
{
"data": {
"values": [ 10, 20, 30 ]
},
"name": "item 2"
},
{
"data": {
"values": [ 1, 2, 3 ]
},
"name": "item 1"
}
]

Order by a calculated column

If the results of connection.operation have the format:

[
{
"a": 4,
"b": 3,
},
{
"a": 2,
"b": 2,
},
{
"a": 2,
"b": 3,
}
]

The query:

SELECT { ...T , key: a * b }
FROM connection.operation AS T
ORDER BY key ASC

Will reorder the rows:

[
{
"a": 2,
"b": 2,
"key": 4
},
{
"a": 2,
"b": 3,
"key": 6
},
{
"a": 4,
"b": 3,
"key": 12
}
]

Limit clause

The limit clause specifies the maximum number of results to return in the query. The syntax is:

LIMIT <number>

External parameters

To access the value of a parameter defined on the SQL operation, use the following syntax:

@<param-name>

The <param-name> has the same restrictions as operations: it must start with an underscore or letter, after which letters, digits, underscores, and dashes are allowed.

External parameters can be used anywhere literal values are allowed, including column selection and the WHERE clause.

Examples

Using parameters in WHERE clause:

In the following query:

SELECT * FROM connection.operation
WHERE input1 = @myParam

The value of the parameter myParam will be passed to input1.

Using parameters with string concatenation:

Because parameters get replaced with their value at runtime, they can be used in binary expressions. In the following query, if myParam is a string, it can be used in string concatenation:

SELECT * FROM connection.operation
WHERE input1 = 'foo ' + @myParam

Using parameters in a JSON template:

The following query again shows a parameter being used in a binary expression, but this time in column selection:

SELECT { foo: @myParam + 1 }

Keywords and escaping

Transposit has a number of keywords and reserved words in SQL. In order to use a reserved word as an identifier in a query, it is necessary to escape it using backticks:

`<keyword>`

Escaping is necessary when the identifier is used in many parts of the query, including column selection, WHERE clauses, and EXPAND BY. However, it is not necessary to escape operation names or external variables.

Examples

In these examples, we will use the reserved word "values" to demonstrate when it is necessary to escape identifiers.

Keywords in column selection

It is necessary to escape column names in the SELECT clause.

SELECT `values` FROM connection.operation

Keywords in the WHERE clause

It is necessary to escape column names on both sides of a condition in a WHERE clause.

SELECT * FROM connection.operation
WHERE `values`=foo
SELECT * FROM connection.operation
WHERE foo=`values`

However, it is not necessary if the keyword is part of a longer path.

SELECT * FROM connection.operation
WHERE foo=path.to.values

Keywords in the EXPAND BY clause

It is necessary to escape column names in the EXPAND BY clause.

SELECT * FROM connection.operation
EXPAND BY `values`

Like with the WHERE clause, it is not necessary to escape if part of a longer path.

SELECT * FROM connection.operation
EXPAND BY path.to.values

Keywords in the connection or operation name

It is not necessary to escape keywords in the connection name.

SELECT * FROM values.operation

Similarly, it is not necessary to escape keywords in the operation name.

SELECT * FROM connection.values

Keywords in external parameters

It is not necessary to escape external parameters, as they are already denoted by the @ symbol.

SELECT * FROM connection.operation
WHERE foo=@values

Keywords in literal strings

It is not necessary to escape keywords when they are used in a string literal.

SELECT * FROM connection.operation
WHERE foo='values'

Comments

Transposit supports block comments using /* */ or single line comments using the -- prefix.

Examples

Line comment

SELECT *  -- the rest of the line is a comment
FROM connection.operation
WHERE foo=bar

Block comment

SELECT * /* 
everything in here is a comment
*/
FROM connection.operation
WHERE foo=bar

Pagination and streaming

Transposit automatically paginates many APIs (see the FAQ). This feature is deeply integrated with our SQL engine, allowing a query to dynamically pull more results from data connectors until it has reached the desired limit.

For example, if we have the following query:

SELECT * FROM connection.operation LIMIT 10

Assume operation is paginated and has a parameter called pageSize to control the number of results from the API. In this case, Transposit is smart enough to set the pageSize to 10 based on the desired limit.

This is easy for a simple query, but imagine a more complex query that filters the results of the API:

SELECT * FROM connection.operation
WHERE foo='bar'
LIMIT 10

The where clause may filter out some of the results of the API call, leaving us with fewer than ten results. When this happens, the query will iterate and fetch the next page of results from the API. This continues until the query has accumulated ten total results, or the API indicates it is done paginating.

The value that gets passed to the paginated operation is determined by the limit and may propagate to subqueries. For instance,

SELECT * FROM
(SELECT * FROM connection.operation)
LIMIT 10

Even though there is no explicit limit in the inner query, the limit of 10 from the outer query is passed down to the subquery and ultimately to the paginated operation.

If the subquery does have an explicit limit, the minimum of the two limits is used. For instance,

SELECT * FROM
(SELECT * FROM connection.operation LIMIT 8)
LIMIT 10

In this case, the limit passed to connection.operation's pageSize parameter is 8.

INSERT, UPDATE and DELETE Statements

In Transposit when you use SQL to execute operations we use SELECT for all queries. This can be a little confusing when the underlying operation modifies or deletes data. Rather than trying to guess the most natural command, we picked SELECT — but provide a helpful conversion from other commands if you happen to try one. SQL is a powerful language for working with APIs, but the concepts don’t always mesh cleanly. SELECT provides a simple mapping on APIs and allows for powerful use of JOIN syntax, not typically available for other SQL commands.

Additional notes:

  • What about mapping HTTP methods to SQL commands? When you think about SQL and APIs it’s tempting to draw a relationship between SQL commands and API HTTP methods: GET is a natural match for SELECT; DELETE for DELETE. POST is probably closest to INSERT and PATCH and PUT map to UPDATE. This demonstrates the first obvious problem: HTTP methods are not always used consistently and what makes sense to the API vendor might surprise the API consumer (and us too!). We might reasonably expect to see a PUT as an INSERT or a POST as an UPDATE. Some APIs use POST to fetch data, a situation where you might expect to use SELECT. There’s no translation that’s obviously right for all APIs and in all circumstances.
  • In the same way that you execute API requests from SQL operations, you can also execute other Transposit operations, and these other operations can execute different APIs based on custom logic — for example, an operation can create or update an item based on an input parameter or based on another API call — and therefore Transposit operations are not mapped to individual HTTP methods.