dataprof_core/
partial.rs

1use crate::{classification::DataType, source::FileFormat};
2
3/// Result of fast schema inference — column names paired with inferred data types.
4#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
5pub struct SchemaResult {
6    /// Columns with their inferred types, in source order: CSV/Parquet follow
7    /// the header/schema, JSON/JSONL follow the first record's field order with
8    /// later-only fields appended in first-seen order.
9    pub columns: Vec<ColumnSchema>,
10    /// How many rows were sampled to infer the schema (0 for Parquet metadata).
11    pub rows_sampled: usize,
12    /// Time taken for inference in milliseconds.
13    pub inference_time_ms: u128,
14    /// `true` when the entire file was consumed or schema was read from
15    /// metadata; `false` when inference stopped at the sample-size cap and
16    /// the schema may not have fully stabilized.
17    pub schema_stable: bool,
18}
19
20/// A single column's name and inferred data type.
21#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
22pub struct ColumnSchema {
23    pub name: String,
24    pub data_type: DataType,
25}
26
27/// Lightweight structural summary for a file.
28///
29/// This intentionally sits between `quick_row_count()` / `infer_schema()` and a
30/// full profile: it reports shape and sampled per-column counters, not quality
31/// scores, pattern detection, recommendations, or raw values.
32#[derive(Debug, Clone, PartialEq, serde::Serialize, serde::Deserialize)]
33pub struct StructureReport {
34    /// Source identifier, currently the file path passed by the caller.
35    pub source: String,
36    /// Detected or explicitly provided file format.
37    pub format: FileFormat,
38    /// Exact or estimated row count for the source.
39    pub row_count: RowCountEstimate,
40    /// Number of rows consumed for the structural column summaries.
41    pub rows_sampled: usize,
42    /// True when the structural pass consumed the whole source.
43    pub source_exhausted: bool,
44    /// True when the structural pass stopped at the row sample cap.
45    pub truncated: bool,
46    /// Human-readable reason for truncation, if any.
47    pub truncation_reason: Option<String>,
48    /// CSV delimiter used for parsing, when applicable.
49    pub delimiter: Option<String>,
50    /// Per-column structural summaries.
51    pub columns: Vec<StructureColumnSummary>,
52    /// Non-fatal caveats, such as estimated row counts.
53    pub warnings: Vec<String>,
54}
55
56/// Sample-derived or metadata-derived structural summary for one column.
57#[derive(Debug, Clone, PartialEq, serde::Serialize, serde::Deserialize)]
58pub struct StructureColumnSummary {
59    pub name: String,
60    pub data_type: DataType,
61    pub total_count: Option<usize>,
62    pub null_count: Option<usize>,
63    pub null_ratio: Option<f64>,
64    pub unique_count: Option<usize>,
65    pub uniqueness_ratio: Option<f64>,
66    pub distinct_count_approximate: Option<bool>,
67    /// `"sample"` for CSV/JSON/JSONL, `"metadata"` for Parquet.
68    pub provenance: String,
69}
70
71/// Result of a quick row count operation.
72#[derive(Debug, Clone, PartialEq, serde::Serialize, serde::Deserialize)]
73pub struct RowCountEstimate {
74    /// The estimated or exact row count.
75    pub count: u64,
76    /// Whether the count is exact or an estimate.
77    pub exact: bool,
78    /// How the count was obtained.
79    pub method: CountMethod,
80    /// Time taken in milliseconds.
81    pub count_time_ms: u128,
82    /// Approximate 1-sigma relative standard error of the estimate, expressed
83    /// as a fraction of `count` (e.g. `0.02` ≈ ±2%). Populated only for
84    /// sampled estimates where a variance across sampling windows is available;
85    /// `None` for exact counts and when it cannot be estimated.
86    #[serde(default)]
87    pub relative_error: Option<f64>,
88}
89
90/// Method used to obtain the row count.
91#[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
92pub enum CountMethod {
93    /// Read from Parquet file footer metadata (exact, zero row reading).
94    ParquetMetadata,
95    /// Full scan of the file (exact).
96    FullScan,
97    /// Sample-based estimation (approximate).
98    Sampling,
99    /// Full scan of a streaming source (no file metadata available).
100    StreamFullScan,
101}
dataprof_core/partial.rs

dataprof_core/
partial.rs
