Filter by mime

This commit is contained in:
2026-03-15 10:20:15 -07:00
parent 8041fc7531
commit 979fbb9b0d
30 changed files with 258 additions and 93 deletions

View File

@@ -7,7 +7,7 @@ use std::{
use tracing::trace;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue, SyncReadBridge},
};
@@ -82,6 +82,7 @@ impl EpubMetaExtractor {
impl ObjectExtractor for EpubMetaExtractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -89,6 +90,10 @@ impl ObjectExtractor for EpubMetaExtractor {
return Ok(None);
}
if !state.ignore_mime && self.item.mime().essence_str() != "application/epub+zip" {
return Ok(None);
}
Ok(self.get_inner().await?.get(name).cloned())
}

View File

@@ -7,7 +7,7 @@ use std::{
use tracing::trace;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue, SyncReadBridge},
};
@@ -92,6 +92,7 @@ fn strip_html(html: &str) -> String {
impl ObjectExtractor for EpubTextExtractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -99,6 +100,10 @@ impl ObjectExtractor for EpubTextExtractor {
return Ok(None);
}
if !state.ignore_mime && self.item.mime().essence_str() != "application/epub+zip" {
return Ok(None);
}
Ok(self.get_inner().await?.get(name).cloned())
}

View File

@@ -8,7 +8,7 @@ mod epub_text;
pub use epub_text::*;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue},
};
@@ -30,13 +30,14 @@ impl EpubExtractor {
impl ObjectExtractor for EpubExtractor {
async fn field(
&self,
state: &ExtractState,
name: &pile_config::Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
match (name.as_str(), args) {
("text", args) => Ok(Some(
self.text
.field(name, args)
.field(state, name, args)
.await
.map(|x| x.unwrap_or(PileValue::Null))?,
)),

View File

@@ -7,7 +7,7 @@ use std::{
use tracing::trace;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue, SyncReadBridge},
};
@@ -87,6 +87,7 @@ fn tag_to_label(tag: &str) -> Option<Label> {
impl ObjectExtractor for ExifExtractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -100,6 +101,10 @@ impl ObjectExtractor for ExifExtractor {
return Ok(None);
}
if !state.ignore_mime && self.item.mime().type_() != mime::IMAGE {
return Ok(None);
}
Ok(self.get_inner().await?.get(name).cloned())
}

View File

@@ -9,7 +9,7 @@ use std::{
use tracing::trace;
use crate::{
extract::traits::{ListExtractor, ObjectExtractor},
extract::traits::{ExtractState, ListExtractor, ObjectExtractor},
value::{Item, PileValue, SyncReadBridge},
};
@@ -46,16 +46,32 @@ impl FlacImagesExtractor {
return Ok(count);
}
fn mime_ok(&self, state: &ExtractState) -> bool {
if state.ignore_mime {
return true;
}
let essence = self.item.mime().essence_str();
essence == "audio/flac" || essence == "audio/x-flac"
}
}
#[async_trait::async_trait]
impl ListExtractor for FlacImagesExtractor {
async fn get<'a>(&'a self, mut idx: usize) -> Result<Option<PileValue>, std::io::Error> {
async fn get(
&self,
state: &ExtractState,
mut idx: usize,
) -> Result<Option<PileValue>, std::io::Error> {
trace!(
key = self.item.key().as_str(),
"Getting index {idx} from FlacImagesExtractor",
);
if !self.mime_ok(state) {
return Ok(None);
}
let key = self.item.key();
let reader = SyncReadBridge::new_current(self.item.read().await?);
let image = tokio::task::spawn_blocking(move || {
@@ -98,7 +114,11 @@ impl ListExtractor for FlacImagesExtractor {
}))
}
async fn len(&self) -> Result<usize, std::io::Error> {
async fn len(&self, state: &ExtractState) -> Result<usize, std::io::Error> {
if !self.mime_ok(state) {
return Ok(0);
}
if let Some(x) = self.cached_count.get() {
return Ok(*x);
}
@@ -178,12 +198,21 @@ impl FlacExtractor {
return Ok(self.output.get_or_init(|| output));
}
fn mime_ok(&self, state: &ExtractState) -> bool {
if state.ignore_mime {
return true;
}
let essence = self.item.mime().essence_str();
essence == "audio/flac" || essence == "audio/x-flac"
}
}
#[async_trait::async_trait]
impl ObjectExtractor for FlacExtractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -191,6 +220,10 @@ impl ObjectExtractor for FlacExtractor {
return Ok(None);
}
if !self.mime_ok(state) {
return Ok(None);
}
if name.as_str() == "images" {
return Ok(Some(self.images.clone()));
}

View File

@@ -1,5 +1,5 @@
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue},
};
use pile_config::Label;
@@ -88,6 +88,7 @@ impl FsExtractor {
impl ObjectExtractor for FsExtractor {
async fn field(
&self,
_state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {

View File

@@ -9,7 +9,7 @@ use std::{
use tracing::trace;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue, SyncReadBridge},
};
@@ -125,6 +125,7 @@ fn frame_id_to_field(id: &str) -> Cow<'static, str> {
impl ObjectExtractor for Id3Extractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -132,6 +133,10 @@ impl ObjectExtractor for Id3Extractor {
return Ok(None);
}
if !state.ignore_mime && self.item.mime().essence_str() != "audio/mpeg" {
return Ok(None);
}
Ok(self.get_inner().await?.get(name).cloned())
}

View File

@@ -26,7 +26,10 @@ mod sidecar;
pub use sidecar::*;
use crate::{
extract::{misc::MapExtractor, traits::ObjectExtractor},
extract::{
misc::MapExtractor,
traits::{ExtractState, ObjectExtractor},
},
value::{Item, PileValue},
};
@@ -82,10 +85,11 @@ impl ItemExtractor {
impl ObjectExtractor for ItemExtractor {
async fn field(
&self,
state: &ExtractState,
name: &pile_config::Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
self.inner.field(name, args).await
self.inner.field(state, name, args).await
}
#[expect(clippy::unwrap_used)]

View File

@@ -14,7 +14,7 @@ mod pdf_text;
pub use pdf_text::*;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue},
};
@@ -40,6 +40,7 @@ impl PdfExtractor {
impl ObjectExtractor for PdfExtractor {
async fn field(
&self,
state: &ExtractState,
name: &pile_config::Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -50,7 +51,7 @@ impl ObjectExtractor for PdfExtractor {
);
match (name.as_str(), args) {
("text", args) => self.text.field(name, args).await,
("text", args) => self.text.field(state, name, args).await,
("meta", None) => Ok(Some(PileValue::ObjectExtractor(self.meta.clone()))),
#[cfg(feature = "pdfium")]
("pages", None) => Ok(Some(PileValue::ListExtractor(self.pages.clone()))),

View File

@@ -9,7 +9,7 @@ use std::{
use tracing::trace;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue, SyncReadBridge},
};
@@ -124,12 +124,18 @@ fn format_date(d: &Date) -> String {
impl ObjectExtractor for PdfMetaExtractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
if args.is_some() {
return Ok(None);
}
if !state.ignore_mime && self.item.mime().essence_str() != "application/pdf" {
return Ok(None);
}
Ok(self.get_inner().await?.get(name).cloned())
}

View File

@@ -7,7 +7,7 @@ use std::{
use tracing::trace;
use crate::{
extract::traits::ListExtractor,
extract::traits::{ExtractState, ListExtractor},
value::{Item, PileValue, SyncReadBridge},
};
@@ -34,12 +34,20 @@ impl PdfPagesExtractor {
#[async_trait::async_trait]
impl ListExtractor for PdfPagesExtractor {
async fn get(&self, idx: usize) -> Result<Option<PileValue>, std::io::Error> {
async fn get(
&self,
state: &ExtractState,
idx: usize,
) -> Result<Option<PileValue>, std::io::Error> {
trace!(
key = self.item.key().as_str(),
"Getting index {idx} from PdfPagesExtractor",
);
if !state.ignore_mime && self.item.mime().essence_str() != "application/pdf" {
return Ok(None);
}
let bytes = self.get_bytes().await?;
let png = tokio::task::spawn_blocking(move || {
let pdfium = Pdfium::default();
@@ -81,7 +89,11 @@ impl ListExtractor for PdfPagesExtractor {
Ok(Some(value))
}
async fn len(&self) -> Result<usize, std::io::Error> {
async fn len(&self, state: &ExtractState) -> Result<usize, std::io::Error> {
if !state.ignore_mime && self.item.mime().essence_str() != "application/pdf" {
return Ok(0);
}
let bytes = self.get_bytes().await?;
let count = tokio::task::spawn_blocking(move || {
let pdfium = Pdfium::default();
@@ -103,10 +115,10 @@ impl ListExtractor for PdfPagesExtractor {
// Override, extracting all pages is very slow,
// and we can't display binary in json anyway
async fn to_json(&self) -> Result<serde_json::Value, std::io::Error> {
async fn to_json(&self, state: &ExtractState) -> Result<serde_json::Value, std::io::Error> {
Ok(serde_json::Value::String(format!(
"<PdfPages ({} pages)>",
self.len().await?
self.len(state).await?
)))
}
}

View File

@@ -9,7 +9,7 @@ use std::{
use tracing::trace;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue, SyncReadBridge},
};
@@ -104,6 +104,7 @@ impl PdfTextExtractor {
impl ObjectExtractor for PdfTextExtractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -111,6 +112,10 @@ impl ObjectExtractor for PdfTextExtractor {
return Ok(None);
}
if !state.ignore_mime && self.item.mime().essence_str() != "application/pdf" {
return Ok(None);
}
Ok(self.get_inner().await?.get(name).cloned())
}

View File

@@ -4,7 +4,7 @@ use tracing::trace;
use super::TomlExtractor;
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{Item, PileValue},
};
@@ -26,6 +26,7 @@ impl SidecarExtractor {
impl ObjectExtractor for SidecarExtractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -39,7 +40,7 @@ impl ObjectExtractor for SidecarExtractor {
.output
.get_or_init(|| self.item.sidecar().map(TomlExtractor::new))
{
Some(x) => Ok(x.field(name, args).await?),
Some(x) => Ok(x.field(state, name, args).await?),
None => Ok(Some(PileValue::Null)),
}
}

View File

@@ -5,7 +5,7 @@ use std::{
};
use crate::{
extract::traits::ObjectExtractor,
extract::traits::{ExtractState, ObjectExtractor},
value::{AsyncReader, Item, PileValue},
};
@@ -64,6 +64,7 @@ impl TomlExtractor {
impl ObjectExtractor for TomlExtractor {
async fn field(
&self,
state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -71,6 +72,10 @@ impl ObjectExtractor for TomlExtractor {
return Ok(None);
}
if !state.ignore_mime && self.item.mime().type_() != mime::TEXT {
return Ok(None);
}
Ok(self.get_inner().await?.get(name).cloned())
}

View File

@@ -1,6 +1,9 @@
use std::sync::Arc;
use crate::{extract::traits::ListExtractor, value::PileValue};
use crate::{
extract::traits::{ExtractState, ListExtractor},
value::PileValue,
};
pub struct ArrayExtractor {
inner: Arc<Vec<PileValue>>,
@@ -14,11 +17,15 @@ impl ArrayExtractor {
#[async_trait::async_trait]
impl ListExtractor for ArrayExtractor {
async fn get(&self, idx: usize) -> Result<Option<PileValue>, std::io::Error> {
async fn get(
&self,
_state: &ExtractState,
idx: usize,
) -> Result<Option<PileValue>, std::io::Error> {
Ok(self.inner.get(idx).cloned())
}
async fn len(&self) -> Result<usize, std::io::Error> {
async fn len(&self, _state: &ExtractState) -> Result<usize, std::io::Error> {
Ok(self.inner.len())
}
}

View File

@@ -1,7 +1,10 @@
use pile_config::Label;
use std::collections::HashMap;
use crate::{extract::traits::ObjectExtractor, value::PileValue};
use crate::{
extract::traits::{ExtractState, ObjectExtractor},
value::PileValue,
};
#[derive(Default)]
pub struct MapExtractor {
@@ -12,6 +15,7 @@ pub struct MapExtractor {
impl ObjectExtractor for MapExtractor {
async fn field(
&self,
_state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {

View File

@@ -1,4 +1,7 @@
use crate::{extract::traits::ListExtractor, value::PileValue};
use crate::{
extract::traits::{ExtractState, ListExtractor},
value::PileValue,
};
#[derive(Default)]
pub struct VecExtractor {
@@ -7,11 +10,15 @@ pub struct VecExtractor {
#[async_trait::async_trait]
impl ListExtractor for VecExtractor {
async fn get(&self, idx: usize) -> Result<Option<PileValue>, std::io::Error> {
async fn get(
&self,
_state: &ExtractState,
idx: usize,
) -> Result<Option<PileValue>, std::io::Error> {
Ok(self.inner.get(idx).cloned())
}
async fn len(&self) -> Result<usize, std::io::Error> {
async fn len(&self, _state: &ExtractState) -> Result<usize, std::io::Error> {
Ok(self.inner.len())
}
}

View File

@@ -2,7 +2,10 @@ use pile_config::Label;
use smartstring::{LazyCompact, SmartString};
use std::sync::Arc;
use crate::{extract::traits::ObjectExtractor, value::PileValue};
use crate::{
extract::traits::{ExtractState, ObjectExtractor},
value::PileValue,
};
pub struct StringExtractor {
item: Arc<SmartString<LazyCompact>>,
@@ -18,6 +21,7 @@ impl StringExtractor {
impl ObjectExtractor for StringExtractor {
async fn field(
&self,
_state: &ExtractState,
name: &Label,
args: Option<&str>,
) -> Result<Option<PileValue>, std::io::Error> {
@@ -89,7 +93,10 @@ mod tests {
#[expect(clippy::unwrap_used)]
async fn field(ext: &StringExtractor, name: &str, args: Option<&str>) -> Option<PileValue> {
ext.field(&Label::new(name).unwrap(), args).await.unwrap()
let state = ExtractState { ignore_mime: false };
ext.field(&state, &Label::new(name).unwrap(), args)
.await
.unwrap()
}
fn string(v: Option<PileValue>) -> Option<String> {

View File

@@ -1,3 +1,13 @@
#[derive(Debug, Clone)]
pub struct ExtractState {
/// If true, extract all fields from all items.
/// Do not pre-filter using mime type.
///
/// This may detect additional fields, but
/// makes extraction take much longer
pub ignore_mime: bool,
}
/// An attachment that extracts metadata from an [Item].
///
/// Metadata is exposed as an immutable map of {label: value},
@@ -15,6 +25,7 @@ pub trait ObjectExtractor: Send + Sync {
/// this fn should return `Ok(Some(None))`.
async fn field(
&self,
state: &ExtractState,
name: &pile_config::Label,
args: Option<&str>,
) -> Result<Option<crate::value::PileValue>, std::io::Error>;
@@ -25,15 +36,15 @@ pub trait ObjectExtractor: Send + Sync {
async fn fields(&self) -> Result<Vec<pile_config::Label>, std::io::Error>;
/// Convert this to a JSON value.
async fn to_json(&self) -> Result<serde_json::Value, std::io::Error> {
async fn to_json(&self, state: &ExtractState) -> Result<serde_json::Value, std::io::Error> {
let keys = self.fields().await?;
let mut map = serde_json::Map::new();
for k in &keys {
let v = match self.field(k, None).await? {
let v = match self.field(state, k, None).await? {
Some(x) => x,
None => continue,
};
map.insert(k.to_string(), Box::pin(v.to_json()).await?);
map.insert(k.to_string(), Box::pin(v.to_json(state)).await?);
}
Ok(serde_json::Value::Object(map))
@@ -49,25 +60,25 @@ pub trait ListExtractor: Send + Sync {
/// Indices start at zero, and must be consecutive.
/// - returns `None` if `idx` is out of range
/// - returns `Some(Null)` if `None` is at `idx`
async fn get(&self, idx: usize) -> Result<Option<crate::value::PileValue>, std::io::Error>;
async fn get(
&self,
state: &ExtractState,
idx: usize,
) -> Result<Option<crate::value::PileValue>, std::io::Error>;
async fn len(&self) -> Result<usize, std::io::Error>;
async fn is_empty(&self) -> Result<bool, std::io::Error> {
Ok(self.len().await? == 0)
}
async fn len(&self, state: &ExtractState) -> Result<usize, std::io::Error>;
/// Convert this list to a JSON value.
async fn to_json(&self) -> Result<serde_json::Value, std::io::Error> {
let len = self.len().await?;
async fn to_json(&self, state: &ExtractState) -> Result<serde_json::Value, std::io::Error> {
let len = self.len(state).await?;
let mut list = Vec::with_capacity(len);
for i in 0..len {
#[expect(clippy::expect_used)]
let v = self
.get(i)
.get(state, i)
.await?
.expect("value must be present according to length");
list.push(Box::pin(v.to_json()).await?);
list.push(Box::pin(v.to_json(state)).await?);
}
Ok(serde_json::Value::Array(list))

View File

@@ -9,7 +9,7 @@ use crate::{
item::ItemExtractor,
misc::{ArrayExtractor, MapExtractor, VecExtractor},
string::StringExtractor,
traits::{ListExtractor, ObjectExtractor},
traits::{ExtractState, ListExtractor, ObjectExtractor},
},
value::Item,
};
@@ -91,7 +91,11 @@ impl PileValue {
}
}
pub async fn query(&self, query: &ObjectPath) -> Result<Option<Self>, std::io::Error> {
pub async fn query(
&self,
state: &ExtractState,
query: &ObjectPath,
) -> Result<Option<Self>, std::io::Error> {
let mut out: Option<PileValue> = Some(self.clone());
for s in &query.segments {
@@ -106,7 +110,7 @@ impl PileValue {
}
};
out = e.field(name, args.as_deref()).await?;
out = e.field(state, name, args.as_deref()).await?;
}
PathSegment::Index(idx) => {
@@ -121,7 +125,7 @@ impl PileValue {
let idx = if *idx >= 0 {
usize::try_from(*idx).ok()
} else {
usize::try_from(e.len().await? as i64 - idx).ok()
usize::try_from(e.len(state).await? as i64 - idx).ok()
};
let idx = match idx {
@@ -132,7 +136,7 @@ impl PileValue {
}
};
out = e.get(idx).await?;
out = e.get(state, idx).await?;
}
}
}
@@ -147,7 +151,10 @@ impl PileValue {
/// - `ObjectExtractor` is recursed into; returns `Some(Object(map))` with
/// only the fields that had data, or `None` if all fields were absent.
/// - `Array` / `ListExtractor` are treated as opaque leaf values (not descended into).
pub async fn count_fields(&self) -> Result<Option<Value>, std::io::Error> {
pub async fn count_fields(
&self,
state: &ExtractState,
) -> Result<Option<Value>, std::io::Error> {
Ok(match self {
Self::Null => None,
@@ -156,18 +163,18 @@ impl PileValue {
}
Self::Array(x) => (!x.is_empty()).then(|| Value::Number(1u64.into())),
Self::ListExtractor(x) => (!x.is_empty().await?).then(|| Value::Number(1u64.into())),
Self::ListExtractor(x) => (x.len(state).await? > 0).then(|| Value::Number(1u64.into())),
Self::ObjectExtractor(_) | Self::Item(_) => {
let e = self.object_extractor();
let keys = e.fields().await?;
let mut map = Map::new();
for k in &keys {
let v = match e.field(k, None).await? {
let v = match e.field(state, k, None).await? {
Some(x) => x,
None => continue,
};
if let Some(counted) = Box::pin(v.count_fields()).await? {
if let Some(counted) = Box::pin(v.count_fields(state)).await? {
map.insert(k.to_string(), counted);
}
}
@@ -187,7 +194,7 @@ impl PileValue {
}
}
pub async fn to_json(&self) -> Result<Value, std::io::Error> {
pub async fn to_json(&self, state: &ExtractState) -> Result<Value, std::io::Error> {
Ok(match self {
Self::Null => Value::Null,
Self::U64(x) => Value::Number((*x).into()),
@@ -201,12 +208,12 @@ impl PileValue {
Self::Array(_) | Self::ListExtractor(_) => {
let e = self.list_extractor();
return e.to_json().await;
return e.to_json(state).await;
}
Self::ObjectExtractor(_) | Self::Item(_) => {
let e = self.object_extractor();
return e.to_json().await;
return e.to_json(state).await;
}
})
}