csift 0.12.2

ripgrep for Claude Code session transcripts: fast regex list/search over ~/.claude/projects/**/*.jsonl
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259

//! search census NUMBERS: the per-axis record counts, the excluded tally, the turn key's
//! transcript qualifier, and the transcript-id cap. Every assertion here is an exact
//! figure or a conservation law - a census that renders the right keys with the wrong
//! counts is the failure mode a `contains` check cannot see.

use crate::harness::*;

const CENSUS_ENC: &str = "-Users-dev-example-project";
const CENSUS_SESS: &str = "00000000-0000-4000-8000-000000000031";

/// One turn with a paired tool round-trip and one ORPHAN tool_result (its `tool_use` is
/// out of scope), plus a second plain turn. Every record carries a label, so a whole-scope
/// census sees a known set: 5 records in turn 0, 2 in turn 1.
fn census_home() -> Home {
    let h = Home::new();
    h.write(
        &format!("{CENSUS_ENC}/{CENSUS_SESS}.jsonl"),
        concat!(
            r#"{"type":"user","uuid":"u0","timestamp":"2026-06-07T05:00:00.000Z","message":{"role":"user","content":"first ask"}}"#, "\n",
            r#"{"type":"assistant","uuid":"a0","parentUuid":"u0","timestamp":"2026-06-07T05:00:01.000Z","message":{"role":"assistant","model":"claude-x-1","content":[{"type":"text","text":"reply one"}]}}"#, "\n",
            r#"{"type":"assistant","uuid":"a1","parentUuid":"a0","timestamp":"2026-06-07T05:00:02.000Z","message":{"role":"assistant","model":"claude-x-1","content":[{"type":"tool_use","id":"t1","name":"Read","input":{"file":"x.rs"}}]}}"#, "\n",
            r#"{"type":"user","uuid":"c1","parentUuid":"a1","timestamp":"2026-06-07T05:00:03.000Z","message":{"role":"user","content":[{"type":"tool_result","tool_use_id":"t1","content":"read ok"}]}}"#, "\n",
            r#"{"type":"user","uuid":"c2","parentUuid":"c1","timestamp":"2026-06-07T05:00:04.000Z","message":{"role":"user","content":[{"type":"tool_result","tool_use_id":"sliced-away","content":"orphan result"}]}}"#, "\n",
            r#"{"type":"user","uuid":"u1","timestamp":"2026-06-07T06:00:00.000Z","message":{"role":"user","content":"second ask"}}"#, "\n",
            r#"{"type":"assistant","uuid":"a2","parentUuid":"u1","timestamp":"2026-06-07T06:00:01.000Z","message":{"role":"assistant","model":"claude-x-1","content":[{"type":"text","text":"reply two"}]}}"#, "\n",
        ),
    );
    h
}

/// Run a whole-scope census on one axis and return (key -> count, summary).
fn census(h: &Home, axis: &str) -> (Vec<(String, u64)>, serde_json::Value) {
    let out = h.run(&[
        "search",
        "",
        &at(CENSUS_SESS),
        "--count-by",
        axis,
        "--format",
        "json",
    ]);
    assert!(out.success, "{axis} census failed: {}", out.stderr);
    let rows = json_rows(&out.stdout, "census")
        .iter()
        .map(|r| {
            (
                r["key"].as_str().expect("a census key").to_string(),
                r["records"].as_u64().expect("a census count"),
            )
        })
        .collect();
    (rows, json_summary(&out.stdout))
}

#[test]
fn every_census_axis_files_each_matched_record_exactly_once() {
    let h = census_home();
    // The conservation law every non-label axis obeys: each matched record lands under
    // exactly one key OR in the excluded tally. A count that never increments still
    // renders its key, so only the arithmetic catches it.
    for axis in ["turn", "session", "model", "pairing", "tool", "result"] {
        let (rows, summary) = census(&h, axis);
        let matched = summary["matched_records"]
            .as_u64()
            .expect("matched_records");
        let excluded = summary["excluded_records"]
            .as_u64()
            .expect("excluded_records");
        assert_eq!(matched, 7, "the fixture has 7 labelled records ({axis})");
        let filed: u64 = rows.iter().map(|(_, n)| n).sum();
        assert_eq!(
            filed + excluded,
            matched,
            "{axis}: {filed} filed + {excluded} excluded must equal {matched} matched, rows {rows:?}"
        );
        assert_eq!(
            rows.len() as u64,
            summary["distinct_keys"].as_u64().expect("distinct_keys"),
            "{axis} rows: {rows:?}"
        );
    }
}

#[test]
fn the_turn_axis_is_a_histogram_of_the_records_each_turn_holds() {
    let h = census_home();
    let (rows, _) = census(&h, "turn");
    assert_eq!(
        rows,
        vec![("t0".to_string(), 5), ("t1".to_string(), 2)],
        "turn 0 holds the opener, the reply, the tool round-trip and the orphan"
    );
}

#[test]
fn the_session_axis_counts_every_record_under_its_transcript() {
    let h = census_home();
    let (rows, _) = census(&h, "session");
    assert_eq!(rows, vec![(CENSUS_SESS.to_string(), 7)]);
}

#[test]
fn the_pairing_axis_separates_the_orphan_from_the_paired_round_trip() {
    let h = census_home();
    let (rows, summary) = census(&h, "pairing");
    assert_eq!(
        rows,
        vec![("paired".to_string(), 2), ("orphan".to_string(), 1)],
        "the use and its result are paired; the result whose use is out of scope is not"
    );
    // The four records outside the axis (no tool_use_id) are reported, never dropped.
    assert_eq!(summary["excluded_records"], serde_json::json!(4));
}

#[test]
fn the_model_axis_counts_the_stamped_records_and_reports_the_rest() {
    let h = census_home();
    let (rows, summary) = census(&h, "model");
    assert_eq!(rows, vec![("claude-x-1".to_string(), 3)]);
    assert_eq!(
        summary["excluded_records"],
        serde_json::json!(4),
        "a record with no model stamp is excluded AND tallied"
    );
}

#[test]
fn the_turn_key_names_its_transcript_only_when_the_scope_spans_several() {
    let h = Home::new();
    let one = "00000000-0000-4000-8000-000000000032";
    let two = "00000000-0000-4000-8000-000000000033";
    for (sess, hour) in [(one, "05"), (two, "06")] {
        h.write(
            &format!("{CENSUS_ENC}/{sess}.jsonl"),
            &format!(
                "{}\n",
                format_args!(
                    r#"{{"type":"user","timestamp":"2026-06-07T{hour}:00:00.000Z","message":{{"role":"user","content":"SPANWORD ask"}}}}"#
                )
            ),
        );
    }
    // Two transcripts in scope: a bare `t0` from each would collide into one row, so the
    // key carries the FULL transcript id (a `@` target as-is).
    let out = h.run(&[
        "search",
        "SPANWORD",
        "--count-by",
        "turn",
        "--format",
        "json",
    ]);
    assert!(out.success, "stderr: {}", out.stderr);
    let keys: Vec<String> = json_rows(&out.stdout, "census")
        .iter()
        .map(|r| r["key"].as_str().unwrap().to_string())
        .collect();
    assert_eq!(
        keys,
        vec![format!("{one}\u{b7}t0"), format!("{two}\u{b7}t0")],
        "a multi-transcript turn census qualifies every key"
    );
    // One transcript in scope: the qualifier would be noise, so it is dropped.
    let solo = h.run(&[
        "search",
        "SPANWORD",
        &at(one),
        "--count-by",
        "turn",
        "--format",
        "json",
    ]);
    let solo_keys: Vec<String> = json_rows(&solo.stdout, "census")
        .iter()
        .map(|r| r["key"].as_str().unwrap().to_string())
        .collect();
    assert_eq!(solo_keys, vec!["t0".to_string()]);
}

#[test]
fn the_transcript_id_list_is_capped_at_a_hundred_with_an_explicit_flag() {
    let h = Home::new();
    let write_n = |n: usize| {
        for i in 0..n {
            let sess = format!("00000000-0000-4000-8000-{i:012}");
            h.write(
                &format!("{CENSUS_ENC}/{sess}.jsonl"),
                &format!(
                    "{}\n",
                    format_args!(
                        r#"{{"type":"user","timestamp":"2026-06-07T05:00:00.000Z","message":{{"role":"user","content":"CAPWORD ask"}}}}"#
                    )
                ),
            );
        }
    };
    // Exactly the cap: every id is listed and nothing was withheld.
    write_n(100);
    let out = h.run(&["search", "CAPWORD", "--format", "json"]);
    assert!(out.success, "stderr: {}", out.stderr);
    let summary = json_summary(&out.stdout);
    assert_eq!(summary["transcript_ids"].as_array().unwrap().len(), 100);
    assert_eq!(
        summary["transcript_ids_truncated"],
        serde_json::json!(false),
        "a hundred ids fit: {summary}"
    );
    // One past it: the list stays a hundred and says so.
    write_n(101);
    let out = h.run(&["search", "CAPWORD", "--format", "json"]);
    let summary = json_summary(&out.stdout);
    assert_eq!(summary["transcript_ids"].as_array().unwrap().len(), 100);
    assert_eq!(
        summary["transcript_ids_truncated"],
        serde_json::json!(true),
        "the hundred-and-first is withheld, never silently: {summary}"
    );
}

#[test]
fn two_merged_sidecar_records_census_as_two_records() {
    // A merged elicitation record has NO physical line, so the census grouper cannot use
    // the line number to tell one record from the next - each line-less hit is its own
    // group. Two pending markers in one turn are two records, not one.
    let h = sidecar_session_home();
    h.write(
        &format!("{ENC}/{SESS}/elicitations.jsonl"),
        &format!(
            "{}\n{}\n",
            mcp_pending_line("el-a", "2026-06-27T01:10:00.000Z", "gdrive", "zzmcp alpha"),
            mcp_pending_line("el-b", "2026-06-27T01:11:00.000Z", "github", "zzmcp beta"),
        ),
    );
    let out = h.run(&[
        "search",
        "zzmcp",
        &at(SESS),
        "--count-by",
        "label",
        "--format",
        "json",
    ]);
    assert!(out.success, "stderr: {}", out.stderr);
    let summary = json_summary(&out.stdout);
    assert_eq!(
        summary["matched_records"],
        serde_json::json!(2),
        "both pending markers count: {summary}"
    );
    let rows = json_rows(&out.stdout, "census");
    assert_eq!(
        rows.iter()
            .find(|r| r["key"] == "agent.tool.use")
            .map(|r| r["records"].clone()),
        Some(serde_json::json!(2)),
        "census rows:\n{}",
        out.stdout
    );
}