Reference / deterministic subset
Query support, with the caveats left in view.
Rulattice translates a bounded set of constructs through one canonical model. This page lists what each adapter reads and writes, how meaning is preserved, and what is rejected. A structural match is not proof that the destination schema carries the same meaning.
Supported constructs
| Construct | KQL | Splunk SPL | Cortex XQL | CQL / LogScale |
|---|---|---|---|---|
| Source | Table | index=x | dataset = x | #event_simpleName=x / #repo=x |
| Relative time | TimeGenerated >= ago(1h), between (ago(1h) .. now()) | earliest=-1h | config timeframe = 1h | setTimeInterval(start=1h), test(@timestamp >= now() - duration("1h")) |
| Boolean logic | and or not() | AND OR NOT, implicit AND | and or not | and or not !, implicit AND |
| Equality | == != =~ !~ | = != (search), = == != (where) | = != | = != |
| Numeric compare | > >= < <= | same | same | test(x > 5) |
| Contains / prefix / suffix | contains startswith endswith (+ _cs), has | "*v*" (search), like(), match() | contains, ~= | "*v*", /v/i, wildcard() |
| Lists | in in~ !in | IN ( ), in() | in ( ) not in | in(field=, values=[ ]) |
| Regex | matches regex | match() | ~= | /re/flags, regex() |
| Fields | project project-away | fields, fields -, table | fields, fields -x | select() table() drop() |
| Aggregation | summarize count() dcount() sum() min() max() avg() by, count | stats count dc sum min max avg BY | comp count() count_distinct() ... by | groupBy(), count(), stats() |
| Computed fields | extend a = expr, b = expr | eval a=expr, b=expr | alter a = expr | a := expr, eval(a = expr), lower()/upper()/length(f, as=), concat([f], as=) |
| Functions | tolower toupper strlen strcat substring coalesce tostring todouble iff case | lower upper len . substr coalesce tostring tonumber if case | lowercase uppercase len concat coalesce to_string to_number if add subtract multiply divide | arithmetic, lower upper length concat |
| Renames | project-rename new = old | rename old AS new | emitted as alter + fields - | rename(field=, as=) |
| Constants | let name = literal; (inlined) | n/a | n/a | n/a |
| Event classes | DeviceProcessEvents..., SecurityEvent 4688, ASIM _Im_* | Sysmon EventCode, Windows 4688 | event_type / event_sub_type / agent_os_type = ENUM.* | #event_simpleName + event_platform / ECS event.* |
| Sort / limit | sort by, order by, take, top | sort, head | sort asc|desc, limit | sort(), head() |
Semantics that are preserved
Case sensitivity. KQL
containsand SPL search are case-insensitive; LogScale field matches are case-sensitive. The model records which one the source used and the output reproduces it, using(?i),/i,=~,lower()orconfig case_sensitiveas needed.Wildcards vs literals. In SPL
whereand KQL,*is a literal character. In SPL search and LogScale it is a wildcard. Literal asterisks are escaped in the output and flagged in review.Time fields.
TimeGenerated,_timeand@timestampare bridged automatically.Aggregate names.
count_(KQL),count(SPL) and_count(LogScale) are tracked symbolically, so a threshold or sort after an aggregation still points at the right column.Computed fields. One function set is shared across dialects. SPL's 1-based
substrbecomes KQL's 0-basedsubstring, SPL.becomesstrcat, andcase()becomes nestedif. Names you create are never schema-mapped. SPL and XQL cannot store a boolean in a field, so boolean results are written throughif()and flagged.Event classes. Cortex
ENUMconstants become the target's native selector for the data source you pick (Defender XDR, SecurityEvent or ASIM for KQL; Sysmon or the Security log for SPL). Anything that widens results, like a dropped OS filter, is flagged as caution.Round-trip check. Every output is parsed again in the target language and compared with the source model. A pass means the translation says the same thing structurally.
Schema mapping (opt-in)
Off by default: field names are preserved and known entity fields are flagged. Turn on "Map fields to target data model" to rename them. Every mapping appears as a review item.
| Entity | KQL (ASIM) | SPL (CIM) | XQL (XDM) | CQL (Falcon) | LogScale (ECS-style) | Caveat |
|---|
Known gaps and rejection behavior
Unsupported input returns no partial output. You get the reason instead.
Joins, unions, subsearches and lookups. Tabular and function
letstatements.Functions outside the shared set (for example
split,replace,parse_json,mvcount),substringinto XQL, and conditional or string-literal expressions into LogScale. Computed columns insideproject.Time bucketing such as
bin(),timechartandbucket(); exact timestamps; snap-to modifiers like-24h@h; week durations.count(field), which counts non-null values differently on each platform. Usecountor distinct count.Free-text search terms outside SPL, existence tests (
field=*), field-to-field comparison, and wildcards with an interior*.Mixed sort directions for LogScale, and more than one aggregation stage per query.
KQL
hasis term-based. It is translated as a substring match and flagged, because the substring version can match more events.