feat: add multiply function and numeric modifiers support to JSON key…

[kazeburo]

User description

… parsing

---

PR Type

Enhancement, Tests, Documentation

---

Description

• Add multiply(factor) to JSON-path modifiers.

• Scale count, group-by, and percentile output.

• Reject invalid or non-finite multiply factors.

• Test and document numeric modifier behavior.

---

Diagram Walkthrough

flowchart LR
  A["JSON path with multiply"] -- "parse factor" --> B["Numeric modifier"]
  B -- "apply before output" --> C["Count and group-by metrics"]
  B -- "apply before output" --> D["Percentile metrics"]

Loading

File Walkthrough

	Relevant files
Enhancement	jsonkey.go Parse and validate numeric multiplication modifiers            cmd/mackerel-plugin-jsonl/jsonkey.go Parse numeric modifiers separately from string modifiers and initializers. Add multiply(factor) parsing with finite-number validation. +49/-15  reader.go Apply numeric modifiers to emitted metrics                              cmd/mackerel-plugin-jsonl/reader.go Store and apply numeric modifiers on aggregator functions. Scale count, group-by, and percentile values before output. Accept floating-point counts for duration calculations. +37/-24
Tests	jsonkey_test.go Test numeric modifier parsing and application                        cmd/mackerel-plugin-jsonl/jsonkey_test.go Add a helper for checking parsed modifier counts. Test multiplication alongside existing modifiers and invalid inputs. +45/-60
Documentation	README.md Document multiplication and aggregator modifier support    README.md Document multiply(factor) and its supported aggregators. Clarify modifier and initializer usage restrictions. +5/-2

---

[github-actions]

PR Reviewer Guide 🔍

(Review updated until commit 4c9fb5f)

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Non-finite output A finite factor can still produce an infinite metric. For example, multiply(1e308) applied to a count of 2 yields +Inf, which is then written as a metric value. Reject or handle non-finite results after multiplication. return func(n float64) float64 { return n * factor }, nil Invalid format splitter.IgnoreEmpties discards empty arguments, so malformed inputs such as multiply(,2) and multiply(2,) are accepted as multiply(2). Reject empty arguments rather than silently changing the configured factor expression. parts, err := s.Split(inner, splitter.TrimSpaces, splitter.UnescapeQuotes, splitter.IgnoreEmpties) if err != nil { return nil, err } if len(parts) != 1 { return nil, fmt.Errorf("invalid multiply() format: %s", fn)

✅ Resolved findings

cmd/mackerel-plugin-jsonl/reader.go:192-193

Numeric Overflow

A finite factor can still overflow when applied to a value. For example, multiply(1e308) with a count of 2 emits +Inf, even if dividing by the duration would produce a finite rate. Handle non-finite results before emitting metrics.

cmd/mackerel-plugin-jsonl/jsonkey.go:140

Infinite Output

A finite factor can still produce an infinite result: multiply(1e308) applied to a count of 2 returns +Inf, which is then emitted as a metric value. Check the result before output so a valid factor cannot produce an invalid metric.

cmd/mackerel-plugin-jsonl/jsonkey.go:135-138

Numeric Overflow

A finite factor can still produce a non-finite metric: multiply(1e308) applied to a count of 2 emits +Inf. Check the result before emitting it, not just the parsed factor.

cmd/mackerel-plugin-jsonl/jsonkey.go:131-135

Invalid Factors

strconv.ParseFloat accepts NaN and Inf, so multiply(NaN) or multiply(Inf) passes validation and emits non-finite metric values. Reject non-finite factors before creating the modifier.

[github-actions]

Benchmark Comparison

Base: main
PR: feat/multiply

BenchmarkMainParse results

goos: linux
goarch: amd64
pkg: github.com/monitoring-forge/mackerel-plugin-jsonl/cmd/mackerel-plugin-jsonl
cpu: AMD EPYC 7763 64-Core Processor                
                             │ base-bench.txt │            pr-bench.txt            │
                             │     sec/op     │   sec/op     vs base               │
MainParse_jsonl-4                33.98m ± 19%   32.04m ± 3%  -5.72% (p=0.000 n=10)
MainParse_jsonl_eachkey-4        41.95m ±  0%   43.77m ± 1%  +4.33% (p=0.000 n=10)
MainParse_parse_and_output-4     35.55m ±  1%   33.37m ± 1%  -6.13% (p=0.000 n=10)
geomean                          37.01m         36.03m       -2.62%

                             │ base-bench.txt │            pr-bench.txt             │
                             │      B/op      │     B/op      vs base               │
MainParse_jsonl-4                4.642Mi ± 0%   4.642Mi ± 0%  +0.00% (p=0.001 n=10)
MainParse_jsonl_eachkey-4        4.642Mi ± 0%   4.642Mi ± 0%  +0.00% (p=0.007 n=10)
MainParse_parse_and_output-4     5.412Mi ± 0%   5.412Mi ± 0%  +0.00% (p=0.014 n=10)
geomean                          4.885Mi        4.886Mi       +0.00%

                             │ base-bench.txt │            pr-bench.txt            │
                             │   allocs/op    │  allocs/op   vs base               │
MainParse_jsonl-4                 100.4k ± 0%   100.4k ± 0%       ~ (p=0.337 n=10)
MainParse_jsonl_eachkey-4         100.4k ± 0%   100.4k ± 0%       ~ (p=0.474 n=10)
MainParse_parse_and_output-4      100.5k ± 0%   100.5k ± 0%       ~ (p=1.000 n=10)
geomean                           100.4k        100.4k       +0.00%

[github-actions]

Persistent review updated to latest commit 6cb309a

[github-actions]

PR Code Suggestions ✨

No code suggestions found for the PR.

[github-actions]

Persistent review updated to latest commit 5847c97

[github-actions]

Persistent review updated to latest commit 389e307

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[github-actions]

Persistent review updated to latest commit 4c9fb5f

[github-actions]

PR Code Suggestions ✨

No code suggestions found for the PR.