Correlate infrastructure prelude candidate classes

This commit is contained in:
Jan Petykiewicz 2026-04-18 14:51:27 -07:00
commit 5219d5a9ce
3 changed files with 187 additions and 0 deletions

View file

@ -2154,6 +2154,9 @@ pub struct SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSummary {
pub dominant_candidate_pattern:
Option<SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePattern>,
#[serde(default)]
pub candidate_pattern_correlations:
Vec<SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePatternCorrelation>,
#[serde(default)]
pub profile_span_correlations:
Vec<SmpSavePlacedStructureDynamicSideBufferNamePreludeProfileSpanCorrelation>,
#[serde(default)]
@ -2185,6 +2188,27 @@ pub struct SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSample {
pub previous_profile_chunk_len_to_next_name_or_end: Option<usize>,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePatternCorrelation {
pub child_count_candidate: u16,
pub child_count_candidate_hex: String,
pub saved_primary_child_byte_candidate: u8,
pub saved_primary_child_byte_candidate_hex: String,
pub row_count: usize,
pub unique_name_pair_count: usize,
pub unique_profile_span_count: usize,
#[serde(default)]
pub dominant_primary_name: Option<String>,
#[serde(default)]
pub dominant_secondary_name: Option<String>,
pub dominant_name_pair_count: usize,
#[serde(default)]
pub dominant_profile_span: Option<usize>,
pub dominant_profile_span_count: usize,
#[serde(default)]
pub sample_rows: Vec<SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSample>,
}
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct SmpSavePlacedStructureDynamicSideBufferNamePreludeProfileSpanCorrelation {
pub previous_profile_chunk_len_to_next_name_or_end: usize,
@ -4484,6 +4508,32 @@ fn build_infrastructure_asset_trace_report(
.unwrap_or_else(|| {
"no dominant post-profile outlier breakdown was available".to_string()
}),
side_buffer
.and_then(|probe| probe.payload_envelope_summary.as_ref())
.and_then(|summary| summary.name_prelude_candidate_summary.as_ref())
.map(|summary| {
format!(
"candidate-pattern correlations now split the remaining prelude classes cleanly too: {:?}",
summary
.candidate_pattern_correlations
.iter()
.map(|entry| format!(
"{}/{} rows={} dominant-name={:?}/{:?} x{} dominant-prev-span={:?} x{}",
entry.child_count_candidate_hex,
entry.saved_primary_child_byte_candidate_hex,
entry.row_count,
entry.dominant_primary_name,
entry.dominant_secondary_name,
entry.dominant_name_pair_count,
entry.dominant_profile_span,
entry.dominant_profile_span_count
))
.collect::<Vec<_>>()
)
})
.unwrap_or_else(|| {
"no candidate-pattern correlation summary was available".to_string()
}),
side_buffer
.and_then(|probe| probe.payload_envelope_summary.as_ref())
.and_then(|summary| summary.name_prelude_candidate_summary.as_ref())
@ -13168,6 +13218,125 @@ fn parse_save_placed_structure_dynamic_side_buffer_probe(
.then_with(|| right_key.cmp(left_key))
})
.map(|(value, count)| (*value, *count));
let mut name_prelude_pattern_groups =
BTreeMap::<(u16, u8), Vec<(usize, Option<String>, Option<String>, Option<usize>)>>::new();
for (
name_tag_relative_offset,
primary_name,
secondary_name,
child_count_candidate,
saved_primary_child_byte_candidate,
previous_span,
) in &name_prelude_candidate_rows
{
name_prelude_pattern_groups
.entry((*child_count_candidate, *saved_primary_child_byte_candidate))
.or_default()
.push((
*name_tag_relative_offset,
primary_name.clone(),
secondary_name.clone(),
*previous_span,
));
}
let candidate_pattern_correlations = name_prelude_pattern_groups
.into_iter()
.map(
|(
(child_count_candidate, saved_primary_child_byte_candidate),
rows,
)| {
let mut name_pair_counts =
BTreeMap::<(Option<String>, Option<String>), usize>::new();
let mut profile_span_counts = BTreeMap::<usize, usize>::new();
for (_, primary_name, secondary_name, previous_span) in &rows {
*name_pair_counts
.entry((primary_name.clone(), secondary_name.clone()))
.or_default() += 1;
if let Some(previous_span) = previous_span {
*profile_span_counts.entry(*previous_span).or_default() += 1;
}
}
let dominant_name_pair = name_pair_counts
.iter()
.max_by(|(left_key, left_count), (right_key, right_count)| {
left_count
.cmp(right_count)
.then_with(|| right_key.cmp(left_key))
})
.map(|((primary_name, secondary_name), count)| {
(primary_name.clone(), secondary_name.clone(), *count)
});
let dominant_profile_span = profile_span_counts
.iter()
.max_by(|(left_key, left_count), (right_key, right_count)| {
left_count
.cmp(right_count)
.then_with(|| right_key.cmp(left_key))
})
.map(|(span, count)| (*span, *count));
SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidatePatternCorrelation {
child_count_candidate,
child_count_candidate_hex: format!("0x{child_count_candidate:04x}"),
saved_primary_child_byte_candidate,
saved_primary_child_byte_candidate_hex: format!(
"0x{saved_primary_child_byte_candidate:02x}"
),
row_count: rows.len(),
unique_name_pair_count: name_pair_counts.len(),
unique_profile_span_count: profile_span_counts.len(),
dominant_primary_name: dominant_name_pair
.as_ref()
.and_then(|(primary_name, _, _)| primary_name.clone()),
dominant_secondary_name: dominant_name_pair
.as_ref()
.and_then(|(_, secondary_name, _)| secondary_name.clone()),
dominant_name_pair_count: dominant_name_pair
.map(|(_, _, count)| count)
.unwrap_or_default(),
dominant_profile_span: dominant_profile_span
.map(|(profile_span, _)| profile_span),
dominant_profile_span_count: dominant_profile_span
.map(|(_, count)| count)
.unwrap_or_default(),
sample_rows: rows
.iter()
.take(8)
.enumerate()
.map(
|(
sample_index,
(
name_tag_relative_offset,
primary_name,
secondary_name,
previous_profile_chunk_len_to_next_name_or_end,
),
)| {
SmpSavePlacedStructureDynamicSideBufferNamePreludeCandidateSample {
sample_index,
name_tag_relative_offset: *name_tag_relative_offset,
primary_name: primary_name.clone(),
secondary_name: secondary_name.clone(),
child_count_candidate,
child_count_candidate_hex: format!(
"0x{child_count_candidate:04x}"
),
saved_primary_child_byte_candidate,
saved_primary_child_byte_candidate_hex: format!(
"0x{saved_primary_child_byte_candidate:02x}"
),
previous_profile_chunk_len_to_next_name_or_end:
*previous_profile_chunk_len_to_next_name_or_end,
}
},
)
.collect(),
}
},
)
.take(8)
.collect::<Vec<_>>();
let mut name_prelude_profile_span_groups =
BTreeMap::<usize, Vec<(u16, u8)>>::new();
for (_, _, _, child_count_candidate, saved_primary_child_byte_candidate, previous_span) in
@ -13272,6 +13441,7 @@ fn parse_save_placed_structure_dynamic_side_buffer_probe(
.map(|(_, count)| count)
.unwrap_or_default(),
dominant_candidate_pattern: dominant_name_prelude_candidate_pattern.clone(),
candidate_pattern_correlations,
profile_span_correlations,
sample_rows: name_prelude_candidate_rows
.iter()
@ -24083,6 +24253,7 @@ mod tests {
count: 110,
},
),
candidate_pattern_correlations: Vec::new(),
profile_span_correlations: Vec::new(),
sample_rows: Vec::new(),
},

View file

@ -2992,6 +2992,14 @@ The low helper strip beneath that shared family is tighter now too: `0x0052ecd0`
`TrackCapST_Cap.3dp / Infrastructure` row with compact prefix `0xff0000ff / 0x0001 / 0xff`.
So the next infrastructure pass should target the `BallastCap` outlier family first instead of
spending time on the already-dominant bridge-section class.
The candidate-pattern classes are explicit now too: `0x0055 / 0x00` is a pure
`BallastCapST_Cap.3dp / Infrastructure` class across `18` rows, always preceded by a zero-length
prior profile span, while `0x0002 / 0xff` is a pure
`BridgeSTWood_Section.3dp / Infrastructure` class across `18` rows with dominant prior profile
span `0x06` (`10` rows). So the next infrastructure pass should split its owner questions the
same way: treat `0x0055 / 0x00` as a `BallastCap`-specific boundary artifact class, and treat
`0x0002 / 0xff` as the likely bridge-specific two-child / clone-path class above
`0x0048a1e0/0x0048dcf0`.
The child loader family is explicit now too: local `.rdata` at `0x005cfd00` proves the
`Infrastructure` child vtable uses the shared tagged callback strip directly, with
`+0x40 = 0x00455fc0`, `+0x48 = 0x00455870`, and `+0x4c = 0x00455930`. So the remaining

View file

@ -143,6 +143,14 @@ Working rule:
`TrackCapST_Cap.3dp / Infrastructure` row with compact prefix `0xff0000ff / 0x0001 / 0xff`.
So the next infrastructure slice should target the `BallastCap` outlier family first, not the
already-dominant bridge-section class.
- The candidate-pattern classes are now explicit across the whole stream too: `0x0055 / 0x00`
is a pure `BallastCapST_Cap.3dp / Infrastructure` class across `18` rows, always preceded by a
zero-length prior profile span, while `0x0002 / 0xff` is a pure
`BridgeSTWood_Section.3dp / Infrastructure` class across `18` rows with dominant prior profile
span `0x06` (`10` rows). So the next infrastructure pass should split its owner questions:
treat `0x0055 / 0x00` as a `BallastCap`-specific boundary artifact class, and treat
`0x0002 / 0xff` as the likely bridge-specific two-child / clone-path class above
`0x0048a1e0/0x0048dcf0`.
- Reconstruct the save-side region record body on top of the newly corrected non-direct tagged
region seam (`0x5209/0x520a/0x520b`, stride hint `0x06`, `Marker09` record stems) now that the
`0x55f3` payload is known to be fully consumed by the embedded profile collection on grounded