Understanding the Luminesce SQL query syntax

Luminesce SQL queries are written in the SQL dialect of SQLite, with the syntactical extensions and limitations described below. See which tools are available to write queries.

Note a query can be written either for a data provider or a direct provider, and there are some syntactical differences between them. In particular, direct providers have an arbitrary syntax that may differ for each.

Summary

Luminesce SQL supports most of the syntax of the SQLite SELECT statement. It does not support data definition language (DDL) syntax for creating or modifying database objects.

• Common table expressions (CTE) are not supported in Luminesce SELECT statements but a re-usable variable syntax is available instead.

• TYPES statements can be used before Luminesce SELECT statements to force type conversion on columns.

• FOR-EACH loops can be emulated using APPLY statements.

• Most types of JOIN clause are supported.

• A WAIT modifier is available to pause query execution.

• A number of common SQL functions are supported, and we've supplemented these with sets of custom functions and statistical functions.

• A number of PRAGMA statements specific to Luminesce are supported.

• Server table variables let you cache query results and reuse them in later queries.

• A number of ways to convert to and from JSON are supported.

• Parameterised queries can be run using :<unique-name>:<default-value>.

• Dates must be encapsulated in # rather than ' characters, so for example #2023-11-15 15:59:59# (though other date formats can also be understood).

• Strings must be encapsulated in ' rather than " characters, so for example PortfolioScope = 'Finbourne-Examples'. Note SQLite interprets a column name encapsulated in " characters as a string literal if it doesn't match a valid column name. For this reason, we recommend encapsulating column names in [] characters. Read more on this in the SQLite documentation.

• Multiple lines can be commented out using /* and */ characters at the beginning and end respectively.

See the full list of valid keywords (both inherited from SQLite and proprietary to FINBOURNE).

For a Some.Provider that is authored by FINBOURNE, the fields (columns) available to return or filter can be determined by running the following query:

select FieldName, DataType, IsMain, IsPrimaryKey, SampleValues, Description
        from Sys.Field
        where TableName = 'Some.Provider' and FieldType = 'Column';

Parameters

Data providers may have parameters that can be used to filter a Luminesce query as part of a standard WHERE clause.

Note: Direct providers do not support parameters, although they do have the WITH syntax that provides much the same functionality.

You must assign values to parameters unambiguously (not under an OR clause), and only with the = operator. For example:

select * from Some.Provider where SomeParameter = 12

For a Some.Provider that is authored by FINBOURNE, available parameters can be determined by running the following query:

select FieldName, DataType, ParamDefaultValue, Description
        from Sys.Field
        where TableName = 'Some.Provider' and FieldType = 'Parameter';

Variables

You can use variables to create and populate arbitrary tables of data that can then be used as part of a Luminesce query result, passed into providers, cached in a variety of ways, exported, and more.

This is similar to a combination of the various statements supported by other SQL engines, for example:

• CREATE [TEMP] TABLE...INSERT INTO statements in PostgreSQL

• DECLARE @t TABLE(…) in Microsoft SQL Server

Server table variables

A server table variable caches the result of a query so you can reuse it in later queries. This saves you retrieving the same data from LUSID each time, which is particularly useful when you’re exploring data or building up a complex query.

For example, the following query retrieves 1,000 holdings from LUSID and caches them as @!holdings:

@!holdings = select * from Lusid.Portfolio.Holding limit 1000;
select * from @!holdings;

Later queries can select from @!holdings directly, which is much faster than retrieving the holdings from LUSID again.

Note the following:

• Only the user who ran the query that created a server table variable can access it.

• A server table variable expires after 8 hours by default. You can use a PRAGMA VariableCacheMins = N; to set a lifetime of between 1 minute and 24 hours.  

  • This will apply to all the server table variables in the query.  

  • The underlying data wil be cleared from persistent storage shortly after this time period elapses.  

  • You will still see mention of this variable in the catalog and code-completion for a few hours afterwards, before all mention of it is removed.  It clearly shows as expired in its description. This is to remind you that it did exist, but has expired and help with repopulating it if needed, by showing you the SQL that created it originally.

• There is a size limit per variable. If the data exceeds it, Luminesce displays a warning and you can’t read the variable in later queries, not will it be shown in the catalog / code-completion at all.  

  • With ?= it simply means calculating the variable each time the query is run and having no cache.

• A new server table variable may take a moment to become available for reuse.

Other ways to cache and share data

If the limits above don’t suit your use case, you can use the Client Data Store (see Sql.Db) to create your own providers backed by a Postgres database. You manage the data’s lifetime and permissions yourself, and you can store larger volumes of data and share it with other users.

Note that the Client Data Store is slower to populate, has a fixed shape that you must update manually if your data changes, and requires macros when you first set up a provider so that Luminesce can retrieve data efficiently.

There are many ways to do this in Luminesce; you can, for example, use Drive.SaveAs and Drive.Csv to write data to, and read from, files in Drive.

Retrieving data only when it’s missing

You can use the ?= assignment operator to reuse cached data if it exists, and retrieve it only if it doesn’t. This operator is available only for server table variables. For example:

@!holdings ?= select * from Lusid.Portfolio.Holding limit 1000;
select * from @!holdings;

Luminesce retrieves data from the underlying providers only if @!holdings:

• Hasn’t been cached yet

• Failed to be cached to due size limits

• Was created by different SQL

• Has expired

In every case, the query returns the data. This is useful in views that you call using CROSS APPLY, because caching part of the work means Luminesce doesn’t have to repeat it hundreds of times.  Due to concurrency of APPLY processing it may still be calculated a handful of times before being able to be reused, however.

To update an existing query, add ! to the variable name and ? before the assignment operator.

Many use cases are simply inappropriate for using any form of cache, for example a query where the data would vary by input parameter (thus potentially caching entirely the wrong data for the next query).  It is up to the query author to be sure a cache is appropriate for the cases when they use it.

Finding server table variables in the Luminesce catalogue

@!tableVariables appear in your catalog alongside other tables, including in the catalogue explorer and code completion helpers.

Luminesce also carries lineage forward, so you can see the SQL that created a server table variable and the full lineage behind that SQL.

Type conversion

You can use the TYPES statement before any SELECT statement in a Luminesce query to force type conversion on columns. Note that this overrides the default SQLite behaviour of flexible typing.

Valid types are BIGINT, INT, DOUBLE, DECIMAL, BOOLEAN, TEXT, DATE, TIME and DATETIME.

FOR-EACH loops

You can use the CROSS APPLY or OUTER APPLY statement to emulate a FOR-EACH loop. This is useful to execute a Luminesce query in parallel on a range of parameter inputs.

For example, the following query uses OUTER APPLY with the Lusid.Portfolio.Holding provider to generate a holdings report for each of the last 3 days:

@days = select date('now', '-' || Value || ' days') as dt
        from Tools.[Range] where number = 3 -- go back N days from today
        and Start = 0; -- subtract 0 days, to include today (1 would exclude today)

        select d.dt, results.* -- outer query
        from @days as d
        outer apply (
        select h.^ -- inner query
        from lusid.portfolio.holding as h
        where h.EffectiveAt = d.dt
        and h.PortfolioScope = 'EMEA'
        and h.PortfolioCode = 'Equities'
        ) as results ;

Note the following:

• Tables used in the inner or outer queries must be aliased, so for example as d in the outer query above and as h in the inner query.

• Columns used in the where clause of the inner query must be explicitly selected in the outer query, so for example d.dt in the inner must be referenced as select d.dt in the outer. Referencing as select d.* would result in an error.

• There must be whitespace before the closing ; character, for example as results ;

• Variables (whether @table or @@scalar) are not expanded in the inner query.

• There is no ON clause for either CROSS APPLY or OUTER APPLY.

JOIN clauses

Luminesce supports the following JOIN clauses: INNER JOIN, LEFT (OUTER) JOIN, CROSS JOIN, FULL (OUTER) JOIN.

Luminesce does not currently support RIGHT (OUTER) JOIN.

Pausing query execution using WAIT

You can use the WAIT modifier immediately after any select * from x statement in a query to wait for code to execute before continuing.

For example, if you create a custom view you may want to wait 5 seconds for it to appear in the Luminesce catalog before querying it:

@model_portfolios_view = use Sys.Admin.SetupView
        --provider=Test.Example.TestHoldings
        ----
        select PortfolioScope, PortfolioCode, BaseCurrency
        from
        Lusid.Portfolio
        where
        PortfolioType = 'Reference'
        enduse;
        --- Create the custom view and then wait 5 seconds
        @result = select * from @model_portfolios_view wait 5;
        --- Query the custom view, having waited for Luminesce to register it
        select * from Test.Example.Testholdings limit 1;

PRAGMAs

You can use one or more of the PRAGMA statements in the following table at any point in a query to override default Luminesce settings for that query. For example, to allow a query to run for 30 minutes before timing out, specify:

PRAGMA names are not case-sensitive. When a query uses a view, some PRAGMAs also apply to the queries inside that view, combined with any the view sets itself: for example, MaxApplyParallelization and MaxParallelVariableCalculations take the lower of the two values, and TimeoutSec, TypeDeterminationRowCount, ProviderRetryNonResponsiveAttempts and MaxJoinFilterKeys take the higher.

pragma TimeoutSec = 1800;

Converting to and from JSON

Luminesce supports the functions exposed by the JSON1 extension library.

You can use the dedicated Tools.JsonExpand provider to emulate the SQLite json.tree function and convert a JSON document into a table of data objects.

To convert a table of data objects into JSON, you can use the SQLite json_group_array and json_object functions. This is required when using a provider that has a field with a Json suffix, for example the PeriodsJson field in the Lusid.Instrument.SimpleCashFlowLoan.Writer provider. The PeriodsJson field is a text field that expects a flattened array of JSON objects as a value, each representing a loan period with PaymentDate, Notional and InterestAmount fields. For example:

@periods = values
        (#2022-03-01#, 100, 5),
        (#2022-06-01#, 90, 4),
        (#2022-09-01#, 50, 4.5),
        (#2022-12-01#, 30, 2.0);

        @@periodsArray = select
        json_group_array(
        json_object(
        'PaymentDate', column1,
        'Notional', column2,
        'InterestAmount', column3
        )
        ) from @periods;

        @table_of_data = select 'MySimpleCashFlowLoan' as DisplayName, 'SimpleCashFlowLoan' as ClientInternal,
        #2022-01-01# as StartDate, #2023-01-01# as MaturityDate, 'GBP' as DomCcy, @@periodsArray as PeriodsJson;

        select * from Lusid.Instrument.SimpleCashFlowLoan.Writer where ToWrite = @table_of_data;

Parameterised queries

Most tools you can use to run Luminesce queries support parameterisation using the following syntax:

• For a single user-specified value: :<unique-name>:<default-value>

• For a fixed list of allowed values: :<unique-name>:{<value-1>, <value-2> [, <value-3>...]}

• For a list of allowed values or a user-specified value (note the *): :<unique-name>:{<value-1>, <value-2> [, <value-3>...] *}

Note

If you want spaces in the unique name, you must encapsulate it in square brackets.

Consider the following example:

select * from Lusid.Portfolio.Txn
        where PortfolioScope = 'Finbourne-Examples'
        and PortfolioCode = :[Choose fund]:{'UK-Equities','US-Equities'} -- User must choose one of these values
        and ShowNonActive = :[Show everything]:True
        and TransactionDate > :After:#2020-07-01#
        and TotalConsideration > :[For more than]:1000
        and TradeCurrency = :[Trade ccy]:{'GBP','USD','EUR' *} -- User can choose one of these values, or specify their own

        -- You can "type" a parameter, yet not provide a default value (the query otherwise derives types from the default value):
        and PortfolioCode != :[For more than 3]:null -- A text parameter with no default, stated explicitly
        and PortfolioCode != :[For more than 4]: -- A text parameter with no default (parameter type defaults to TEXT)
        and TotalConsideration > :[For more than 2]:INT -- A number parameter with no default
        
        -- You can use queries to define options also, for example:
        --   Only show distinct results
        --   Only select one column
        and TradeCurrency = :[Trade ccy 2]:{(select ISOCurrencyCode from CoppClark.Currencies order by 1)*} 
                            -- Currencies without a default, any text allowed
        and TradeCurrency = :[Trade ccy 4]:{'USD' (select distinct ISOCurrencyCode from CoppClark.Currencies order by 1)*} 
                            -- Currencies with a default, any text allowed
        and TradeCurrency = :[Trade ccy 3]:{'USD' (select distinct ISOCurrencyCode from CoppClark.Currencies order by 1 limit 5)} 
                            -- Currencies with a default, only items in the list are allowed
        and TotalConsideration > :[For more than 5]:{INT (select Value from [Tools.Range] where Start = 100 and Number = 100 and Step = 5)}
                            -- No default value, a number value from only items in the list

When this query is run from either the Data Virtualisation > Query Editor or Saved Queries screen in the LUSID web app, the user is prompted to enter the following values at run time:

Note if the tool you use to run a query does not support parameterisation, the query uses the default value.