High cyclomatic complexity
What a high-cyclomatic-complexity finding of fallow health means, why it matters and how to fix it.
What it reports
Function has high cyclomatic complexity.
Health explains how fallow finds it.
Why it matters
McCabe cyclomatic complexity exceeds the configured threshold. Cyclomatic complexity counts the number of independent paths through a function (1 + decision points: if/else, switch cases, loops, ternary, logical operators). High values indicate functions that are hard to test exhaustively. fallow also emits this rule on synthetic <template> findings across every supported template dialect (Angular .html templates and inline @Component({ template: ... }) literals, Vue SFC <template> blocks, Svelte markup, Astro markup), counting each dialect's control-flow blocks (Angular @if/@else if/@for/@case/@defer (when ...) and legacy *ngIf/*ngFor; Vue v-if/v-else-if/v-for; Svelte {#if}/{:else if}/{#each}/{#await}/{#key}; the Astro equivalents) plus ternary and logical operators inside bound attributes and interpolations; on each top-level Svelte {#snippet name(...)} block, emitted as its own <snippet:NAME> unit; and on synthetic <component> rollup findings whose cyclomatic is the worst class method's score plus the template's. Ranking and --targets use the rollup total; JSON exposes the per-half breakdown under component_rollup.
Example
function route(req) has a switch with 14 case labels and two if checks. Its cyclomatic complexity is 17, while health.maxCyclomatic is 15.
Fix
For function findings, extract named helpers, split independent branches, flatten guard clauses, and add tests around the behavior before refactoring. For <template> findings, flatten nested control flow and move branch selection out of the markup: Angular @if/@for/@switch/@defer and legacy *ngIf/*ngFor into computed signals on the component class, Vue v-if/v-else-if/v-for into computed properties, Svelte {#if}/{:else if}/{#each}/{#await}/{#key} into derived state, Astro control flow into props resolved in the frontmatter. In Svelte, moving a repeated or deeply nested block into a top-level {#snippet} gives it its own <snippet:NAME> unit with nesting rebased to zero, so the in-file extraction moves the score; splitting a large template into child components lowers the score in every dialect. For <component> rollup findings, attack the larger half first; the per-half breakdown lives in component_rollup. To report these findings without a failed run, set the complexity-cyclomatic or complexity-cognitive rule to warn.
Facts
| Fact | Value |
|---|---|
| Issue type | high-cyclomatic-complexity |
| Rule id in SARIF output | fallow/high-cyclomatic-complexity |
| Command | fallow health |
| Default severity | error |
Config key in rules | complexity-cyclomatic |
| Filter option | --complexity |
| Suppression comment | // fallow-ignore-next-line complexity |
Fixable with fallow fix | No |
| Opt-in | No |
When to keep the code
- Generated code, such as a GraphQL or OpenAPI client: leave it out with
health.ignore. - A test file with many
it()blocks can have a high count. Tests must cover many paths. - To keep one function visible with a higher limit, add a
health.thresholdOverridesentry. A suppression hides the function completely.