pub struct Store { /* private fields */ }Expand description
A database for reading and writing objects to disk, one file per object.
Implementations§
source§impl Store
impl Store
Object lookup
sourcepub fn contains(&self, id: impl AsRef<oid>) -> bool
pub fn contains(&self, id: impl AsRef<oid>) -> bool
Returns true if the given id is contained in our repository.
Examples found in repository?
src/store_impls/dynamic/header.rs (line 162)
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
fn try_header_inner<'b>(
&'b self,
mut id: &'b git_hash::oid,
snapshot: &mut load_index::Snapshot,
recursion: Option<DeltaBaseRecursion<'_>>,
) -> Result<Option<Header>, Error> {
if let Some(r) = recursion {
if r.depth >= self.max_recursion_depth {
return Err(Error::DeltaBaseRecursionLimit {
max_depth: self.max_recursion_depth,
id: r.original_id.to_owned(),
});
}
} else if !self.ignore_replacements {
if let Ok(pos) = self
.store
.replacements
.binary_search_by(|(map_this, _)| map_this.as_ref().cmp(id))
{
id = self.store.replacements[pos].1.as_ref();
}
}
'outer: loop {
{
let marker = snapshot.marker;
for (idx, index) in snapshot.indices.iter_mut().enumerate() {
if let Some(handle::index_lookup::Outcome {
object_index: handle::IndexForObjectInPack { pack_id, pack_offset },
index_file,
pack: possibly_pack,
}) = index.lookup(id)
{
let pack = match possibly_pack {
Some(pack) => pack,
None => match self.store.load_pack(pack_id, marker)? {
Some(pack) => {
*possibly_pack = Some(pack);
possibly_pack.as_deref().expect("just put it in")
}
None => {
// The pack wasn't available anymore so we are supposed to try another round with a fresh index
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
*snapshot = new_snapshot;
self.clear_cache();
continue 'outer;
}
None => {
// nothing new in the index, kind of unexpected to not have a pack but to also
// to have no new index yet. We set the new index before removing any slots, so
// this should be observable.
return Ok(None);
}
}
}
},
};
let entry = pack.entry(pack_offset);
let res = match pack.decode_header(entry, |id| {
index_file.pack_offset_by_id(id).map(|pack_offset| {
git_pack::data::decode::header::ResolvedBase::InPack(pack.entry(pack_offset))
})
}) {
Ok(header) => Ok(header.into()),
Err(git_pack::data::decode::Error::DeltaBaseUnresolved(base_id)) => {
// Only with multi-pack indices it's allowed to jump to refer to other packs within this
// multi-pack. Otherwise this would constitute a thin pack which is only allowed in transit.
// However, if we somehow end up with that, we will resolve it safely, even though we could
// avoid handling this case and error instead.
let hdr = self
.try_header_inner(
&base_id,
snapshot,
recursion
.map(|r| r.inc_depth())
.or_else(|| DeltaBaseRecursion::new(id).into()),
)
.map_err(|err| Error::DeltaBaseLookup {
err: Box::new(err),
base_id,
id: id.to_owned(),
})?
.ok_or_else(|| Error::DeltaBaseMissing {
base_id,
id: id.to_owned(),
})?;
let handle::index_lookup::Outcome {
object_index:
handle::IndexForObjectInPack {
pack_id: _,
pack_offset,
},
index_file,
pack: possibly_pack,
} = match snapshot.indices[idx].lookup(id) {
Some(res) => res,
None => {
let mut out = None;
for index in snapshot.indices.iter_mut() {
out = index.lookup(id);
if out.is_some() {
break;
}
}
out.unwrap_or_else(|| {
panic!("could not find object {} in any index after looking up one of its base objects {}", id, base_id)
})
}
};
let pack = possibly_pack
.as_ref()
.expect("pack to still be available like just now");
let entry = pack.entry(pack_offset);
pack.decode_header(entry, |id| {
index_file
.pack_offset_by_id(id)
.map(|pack_offset| {
git_pack::data::decode::header::ResolvedBase::InPack(
pack.entry(pack_offset),
)
})
.or_else(|| {
(id == base_id).then(|| {
git_pack::data::decode::header::ResolvedBase::OutOfPack {
kind: hdr.kind(),
num_deltas: hdr.num_deltas(),
}
})
})
})
.map(Into::into)
}
Err(err) => Err(err),
}?;
if idx != 0 {
snapshot.indices.swap(0, idx);
}
return Ok(Some(res));
}
}
}
for lodb in snapshot.loose_dbs.iter() {
// TODO: remove this double-lookup once the borrow checker allows it.
if lodb.contains(id) {
return lodb.try_header(id).map(|opt| opt.map(Into::into)).map_err(Into::into);
}
}
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
*snapshot = new_snapshot;
self.clear_cache();
}
None => return Ok(None),
}
}
}More examples
src/store_impls/dynamic/find.rs (line 282)
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340
fn try_find_cached_inner<'a, 'b>(
&'b self,
mut id: &'b git_hash::oid,
buffer: &'a mut Vec<u8>,
pack_cache: &mut impl DecodeEntry,
snapshot: &mut load_index::Snapshot,
recursion: Option<error::DeltaBaseRecursion<'_>>,
) -> Result<Option<(git_object::Data<'a>, Option<git_pack::data::entry::Location>)>, Error> {
if let Some(r) = recursion {
if r.depth >= self.max_recursion_depth {
return Err(Error::DeltaBaseRecursionLimit {
max_depth: self.max_recursion_depth,
id: r.original_id.to_owned(),
});
}
} else if !self.ignore_replacements {
if let Ok(pos) = self
.store
.replacements
.binary_search_by(|(map_this, _)| map_this.as_ref().cmp(id))
{
id = self.store.replacements[pos].1.as_ref();
}
}
'outer: loop {
{
let marker = snapshot.marker;
for (idx, index) in snapshot.indices.iter_mut().enumerate() {
if let Some(handle::index_lookup::Outcome {
object_index: handle::IndexForObjectInPack { pack_id, pack_offset },
index_file,
pack: possibly_pack,
}) = index.lookup(id)
{
let pack = match possibly_pack {
Some(pack) => pack,
None => match self.store.load_pack(pack_id, marker)? {
Some(pack) => {
*possibly_pack = Some(pack);
possibly_pack.as_deref().expect("just put it in")
}
None => {
// The pack wasn't available anymore so we are supposed to try another round with a fresh index
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
*snapshot = new_snapshot;
self.clear_cache();
continue 'outer;
}
None => {
// nothing new in the index, kind of unexpected to not have a pack but to also
// to have no new index yet. We set the new index before removing any slots, so
// this should be observable.
return Ok(None);
}
}
}
},
};
let entry = pack.entry(pack_offset);
let header_size = entry.header_size();
let res = match pack.decode_entry(
entry,
buffer,
|id, _out| {
index_file.pack_offset_by_id(id).map(|pack_offset| {
git_pack::data::decode::entry::ResolvedBase::InPack(pack.entry(pack_offset))
})
},
pack_cache,
) {
Ok(r) => Ok((
git_object::Data {
kind: r.kind,
data: buffer.as_slice(),
},
Some(git_pack::data::entry::Location {
pack_id: pack.id,
pack_offset,
entry_size: r.compressed_size + header_size,
}),
)),
Err(git_pack::data::decode::Error::DeltaBaseUnresolved(base_id)) => {
// Only with multi-pack indices it's allowed to jump to refer to other packs within this
// multi-pack. Otherwise this would constitute a thin pack which is only allowed in transit.
// However, if we somehow end up with that, we will resolve it safely, even though we could
// avoid handling this case and error instead.
// Since this is a special case, we just allocate here to make it work. It's an actual delta-ref object
// which is sent by some servers that points to an object outside of the pack we are looking
// at right now. With the complexities of loading packs, we go into recursion here. Git itself
// doesn't do a cycle check, and we won't either but limit the recursive depth.
// The whole ordeal isn't as efficient as it could be due to memory allocation and
// later mem-copying when trying again.
let mut buf = Vec::new();
let obj_kind = self
.try_find_cached_inner(
&base_id,
&mut buf,
pack_cache,
snapshot,
recursion
.map(|r| r.inc_depth())
.or_else(|| error::DeltaBaseRecursion::new(id).into()),
)
.map_err(|err| Error::DeltaBaseLookup {
err: Box::new(err),
base_id,
id: id.to_owned(),
})?
.ok_or_else(|| Error::DeltaBaseMissing {
base_id,
id: id.to_owned(),
})?
.0
.kind;
let handle::index_lookup::Outcome {
object_index:
handle::IndexForObjectInPack {
pack_id: _,
pack_offset,
},
index_file,
pack: possibly_pack,
} = match snapshot.indices[idx].lookup(id) {
Some(res) => res,
None => {
let mut out = None;
for index in snapshot.indices.iter_mut() {
out = index.lookup(id);
if out.is_some() {
break;
}
}
out.unwrap_or_else(|| {
panic!("could not find object {} in any index after looking up one of its base objects {}", id, base_id)
})
}
};
let pack = possibly_pack
.as_ref()
.expect("pack to still be available like just now");
let entry = pack.entry(pack_offset);
let header_size = entry.header_size();
pack.decode_entry(
entry,
buffer,
|id, out| {
index_file
.pack_offset_by_id(id)
.map(|pack_offset| {
git_pack::data::decode::entry::ResolvedBase::InPack(
pack.entry(pack_offset),
)
})
.or_else(|| {
(id == base_id).then(|| {
out.resize(buf.len(), 0);
out.copy_from_slice(buf.as_slice());
git_pack::data::decode::entry::ResolvedBase::OutOfPack {
kind: obj_kind,
end: out.len(),
}
})
})
},
pack_cache,
)
.map(move |r| {
(
git_object::Data {
kind: r.kind,
data: buffer.as_slice(),
},
Some(git_pack::data::entry::Location {
pack_id: pack.id,
pack_offset,
entry_size: r.compressed_size + header_size,
}),
)
})
}
Err(err) => Err(err),
}?;
if idx != 0 {
snapshot.indices.swap(0, idx);
}
return Ok(Some(res));
}
}
}
for lodb in snapshot.loose_dbs.iter() {
// TODO: remove this double-lookup once the borrow checker allows it.
if lodb.contains(id) {
return lodb
.try_find(id, buffer)
.map(|obj| obj.map(|obj| (obj, None)))
.map_err(Into::into);
}
}
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
*snapshot = new_snapshot;
self.clear_cache();
}
None => return Ok(None),
}
}
}
pub(crate) fn clear_cache(&self) {
self.packed_object_count.borrow_mut().take();
}
}
impl<S> git_pack::Find for super::Handle<S>
where
S: Deref<Target = super::Store> + Clone,
{
type Error = Error;
// TODO: probably make this method fallible, but that would mean its own error type.
fn contains(&self, id: impl AsRef<git_hash::oid>) -> bool {
let id = id.as_ref();
let mut snapshot = self.snapshot.borrow_mut();
loop {
for (idx, index) in snapshot.indices.iter().enumerate() {
if index.contains(id) {
if idx != 0 {
snapshot.indices.swap(0, idx);
}
return true;
}
}
for lodb in snapshot.loose_dbs.iter() {
if lodb.contains(id) {
return true;
}
}
match self.store.load_one_index(self.refresh, snapshot.marker) {
Ok(Some(new_snapshot)) => {
*snapshot = new_snapshot;
self.clear_cache();
}
Ok(None) => return false, // nothing more to load, or our refresh mode doesn't allow disk refreshes
Err(_) => return false, // something went wrong, nothing we can communicate here with this trait. TODO: Maybe that should change?
}
}
}sourcepub fn lookup_prefix(
&self,
prefix: Prefix,
candidates: Option<&mut HashSet<ObjectId>>
) -> Result<Option<Outcome>, Error>
pub fn lookup_prefix(
&self,
prefix: Prefix,
candidates: Option<&mut HashSet<ObjectId>>
) -> Result<Option<Outcome>, Error>
Given a prefix, find an object that matches it uniquely within this loose object
database as Ok(Some(Ok(<oid>))).
If there is more than one object matching the object Ok(Some(Err(())) is returned.
Finally, if no object matches, the return value is Ok(None).
The outer Result is to indicate errors during file system traversal.
Pass candidates to obtain the set of all object ids matching prefix, with the same return value as
one would have received if it remained None.
Examples found in repository?
src/store_impls/dynamic/prefix.rs (line 158)
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
pub fn lookup_prefix(
&self,
prefix: git_hash::Prefix,
mut candidates: Option<&mut HashSet<git_hash::ObjectId>>,
) -> Result<Option<lookup::Outcome>, lookup::Error> {
let mut candidate: Option<git_hash::ObjectId> = None;
loop {
let snapshot = self.snapshot.borrow();
for index in snapshot.indices.iter() {
#[allow(clippy::needless_option_as_deref)] // needed as it's the equivalent of a reborrow.
let lookup_result = index.lookup_prefix(prefix, candidates.as_deref_mut());
if candidates.is_none() && !check_candidate(lookup_result, &mut candidate) {
return Ok(Some(Err(())));
}
}
for lodb in snapshot.loose_dbs.iter() {
#[allow(clippy::needless_option_as_deref)] // needed as it's the equivalent of a reborrow.
let lookup_result = lodb.lookup_prefix(prefix, candidates.as_deref_mut())?;
if candidates.is_none() && !check_candidate(lookup_result, &mut candidate) {
return Ok(Some(Err(())));
}
}
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
drop(snapshot);
*self.snapshot.borrow_mut() = new_snapshot;
}
None => {
return match &candidates {
Some(candidates) => match candidates.len() {
0 => Ok(None),
1 => Ok(candidates.iter().cloned().next().map(Ok)),
_ => Ok(Some(Err(()))),
},
None => Ok(candidate.map(Ok)),
};
}
}
}
fn check_candidate(lookup_result: Option<lookup::Outcome>, candidate: &mut Option<git_hash::ObjectId>) -> bool {
match (lookup_result, &*candidate) {
(Some(Ok(oid)), Some(candidate)) if *candidate != oid => false,
(Some(Ok(_)), Some(_)) | (None, None) | (None, Some(_)) => true,
(Some(Err(())), _) => false,
(Some(Ok(oid)), None) => {
*candidate = Some(oid);
true
}
}
}
}sourcepub fn try_find<'a>(
&self,
id: impl AsRef<oid>,
out: &'a mut Vec<u8>
) -> Result<Option<Data<'a>>, Error>
pub fn try_find<'a>(
&self,
id: impl AsRef<oid>,
out: &'a mut Vec<u8>
) -> Result<Option<Data<'a>>, Error>
Return the object identified by the given ObjectId if present in this database,
writing its raw data into the given out buffer.
Returns Err if there was an error locating or reading the object. Returns Ok<None> if
there was no such object.
Examples found in repository?
src/store_impls/loose/verify.rs (line 59)
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
pub fn verify_integrity(
&self,
mut progress: impl Progress,
should_interrupt: &AtomicBool,
) -> Result<integrity::Statistics, integrity::Error> {
let mut buf = Vec::new();
let sink = crate::sink(self.object_hash);
let mut num_objects = 0;
let start = Instant::now();
let mut progress = progress.add_child_with_id("Validating", *b"VILO"); /* Verify Integrity Loose Objects */
progress.init(None, git_features::progress::count("loose objects"));
for id in self.iter().filter_map(Result::ok) {
let object = self
.try_find(id, &mut buf)
.map_err(|_| integrity::Error::Retry)?
.ok_or(integrity::Error::Retry)?;
let actual_id = sink.write_buf(object.kind, object.data).expect("sink never fails");
if actual_id != id {
return Err(integrity::Error::ObjectHashMismatch {
kind: object.kind,
actual: actual_id,
expected: id,
});
}
object.decode().map_err(|err| integrity::Error::ObjectDecode {
source: err,
kind: object.kind,
id,
})?;
progress.inc();
num_objects += 1;
if should_interrupt.load(Ordering::SeqCst) {
return Err(integrity::Error::Interrupted);
}
}
progress.show_throughput(start);
Ok(integrity::Statistics { num_objects })
}More examples
src/store_impls/dynamic/find.rs (line 284)
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298
fn try_find_cached_inner<'a, 'b>(
&'b self,
mut id: &'b git_hash::oid,
buffer: &'a mut Vec<u8>,
pack_cache: &mut impl DecodeEntry,
snapshot: &mut load_index::Snapshot,
recursion: Option<error::DeltaBaseRecursion<'_>>,
) -> Result<Option<(git_object::Data<'a>, Option<git_pack::data::entry::Location>)>, Error> {
if let Some(r) = recursion {
if r.depth >= self.max_recursion_depth {
return Err(Error::DeltaBaseRecursionLimit {
max_depth: self.max_recursion_depth,
id: r.original_id.to_owned(),
});
}
} else if !self.ignore_replacements {
if let Ok(pos) = self
.store
.replacements
.binary_search_by(|(map_this, _)| map_this.as_ref().cmp(id))
{
id = self.store.replacements[pos].1.as_ref();
}
}
'outer: loop {
{
let marker = snapshot.marker;
for (idx, index) in snapshot.indices.iter_mut().enumerate() {
if let Some(handle::index_lookup::Outcome {
object_index: handle::IndexForObjectInPack { pack_id, pack_offset },
index_file,
pack: possibly_pack,
}) = index.lookup(id)
{
let pack = match possibly_pack {
Some(pack) => pack,
None => match self.store.load_pack(pack_id, marker)? {
Some(pack) => {
*possibly_pack = Some(pack);
possibly_pack.as_deref().expect("just put it in")
}
None => {
// The pack wasn't available anymore so we are supposed to try another round with a fresh index
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
*snapshot = new_snapshot;
self.clear_cache();
continue 'outer;
}
None => {
// nothing new in the index, kind of unexpected to not have a pack but to also
// to have no new index yet. We set the new index before removing any slots, so
// this should be observable.
return Ok(None);
}
}
}
},
};
let entry = pack.entry(pack_offset);
let header_size = entry.header_size();
let res = match pack.decode_entry(
entry,
buffer,
|id, _out| {
index_file.pack_offset_by_id(id).map(|pack_offset| {
git_pack::data::decode::entry::ResolvedBase::InPack(pack.entry(pack_offset))
})
},
pack_cache,
) {
Ok(r) => Ok((
git_object::Data {
kind: r.kind,
data: buffer.as_slice(),
},
Some(git_pack::data::entry::Location {
pack_id: pack.id,
pack_offset,
entry_size: r.compressed_size + header_size,
}),
)),
Err(git_pack::data::decode::Error::DeltaBaseUnresolved(base_id)) => {
// Only with multi-pack indices it's allowed to jump to refer to other packs within this
// multi-pack. Otherwise this would constitute a thin pack which is only allowed in transit.
// However, if we somehow end up with that, we will resolve it safely, even though we could
// avoid handling this case and error instead.
// Since this is a special case, we just allocate here to make it work. It's an actual delta-ref object
// which is sent by some servers that points to an object outside of the pack we are looking
// at right now. With the complexities of loading packs, we go into recursion here. Git itself
// doesn't do a cycle check, and we won't either but limit the recursive depth.
// The whole ordeal isn't as efficient as it could be due to memory allocation and
// later mem-copying when trying again.
let mut buf = Vec::new();
let obj_kind = self
.try_find_cached_inner(
&base_id,
&mut buf,
pack_cache,
snapshot,
recursion
.map(|r| r.inc_depth())
.or_else(|| error::DeltaBaseRecursion::new(id).into()),
)
.map_err(|err| Error::DeltaBaseLookup {
err: Box::new(err),
base_id,
id: id.to_owned(),
})?
.ok_or_else(|| Error::DeltaBaseMissing {
base_id,
id: id.to_owned(),
})?
.0
.kind;
let handle::index_lookup::Outcome {
object_index:
handle::IndexForObjectInPack {
pack_id: _,
pack_offset,
},
index_file,
pack: possibly_pack,
} = match snapshot.indices[idx].lookup(id) {
Some(res) => res,
None => {
let mut out = None;
for index in snapshot.indices.iter_mut() {
out = index.lookup(id);
if out.is_some() {
break;
}
}
out.unwrap_or_else(|| {
panic!("could not find object {} in any index after looking up one of its base objects {}", id, base_id)
})
}
};
let pack = possibly_pack
.as_ref()
.expect("pack to still be available like just now");
let entry = pack.entry(pack_offset);
let header_size = entry.header_size();
pack.decode_entry(
entry,
buffer,
|id, out| {
index_file
.pack_offset_by_id(id)
.map(|pack_offset| {
git_pack::data::decode::entry::ResolvedBase::InPack(
pack.entry(pack_offset),
)
})
.or_else(|| {
(id == base_id).then(|| {
out.resize(buf.len(), 0);
out.copy_from_slice(buf.as_slice());
git_pack::data::decode::entry::ResolvedBase::OutOfPack {
kind: obj_kind,
end: out.len(),
}
})
})
},
pack_cache,
)
.map(move |r| {
(
git_object::Data {
kind: r.kind,
data: buffer.as_slice(),
},
Some(git_pack::data::entry::Location {
pack_id: pack.id,
pack_offset,
entry_size: r.compressed_size + header_size,
}),
)
})
}
Err(err) => Err(err),
}?;
if idx != 0 {
snapshot.indices.swap(0, idx);
}
return Ok(Some(res));
}
}
}
for lodb in snapshot.loose_dbs.iter() {
// TODO: remove this double-lookup once the borrow checker allows it.
if lodb.contains(id) {
return lodb
.try_find(id, buffer)
.map(|obj| obj.map(|obj| (obj, None)))
.map_err(Into::into);
}
}
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
*snapshot = new_snapshot;
self.clear_cache();
}
None => return Ok(None),
}
}
}sourcepub fn try_header(
&self,
id: impl AsRef<oid>
) -> Result<Option<(usize, Kind)>, Error>
pub fn try_header(
&self,
id: impl AsRef<oid>
) -> Result<Option<(usize, Kind)>, Error>
Return only the decompressed size of the object and its kind without fully reading it into memory as tuple of (size, kind).
Returns None if id does not exist in the database.
Examples found in repository?
src/store_impls/dynamic/header.rs (line 163)
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
fn try_header_inner<'b>(
&'b self,
mut id: &'b git_hash::oid,
snapshot: &mut load_index::Snapshot,
recursion: Option<DeltaBaseRecursion<'_>>,
) -> Result<Option<Header>, Error> {
if let Some(r) = recursion {
if r.depth >= self.max_recursion_depth {
return Err(Error::DeltaBaseRecursionLimit {
max_depth: self.max_recursion_depth,
id: r.original_id.to_owned(),
});
}
} else if !self.ignore_replacements {
if let Ok(pos) = self
.store
.replacements
.binary_search_by(|(map_this, _)| map_this.as_ref().cmp(id))
{
id = self.store.replacements[pos].1.as_ref();
}
}
'outer: loop {
{
let marker = snapshot.marker;
for (idx, index) in snapshot.indices.iter_mut().enumerate() {
if let Some(handle::index_lookup::Outcome {
object_index: handle::IndexForObjectInPack { pack_id, pack_offset },
index_file,
pack: possibly_pack,
}) = index.lookup(id)
{
let pack = match possibly_pack {
Some(pack) => pack,
None => match self.store.load_pack(pack_id, marker)? {
Some(pack) => {
*possibly_pack = Some(pack);
possibly_pack.as_deref().expect("just put it in")
}
None => {
// The pack wasn't available anymore so we are supposed to try another round with a fresh index
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
*snapshot = new_snapshot;
self.clear_cache();
continue 'outer;
}
None => {
// nothing new in the index, kind of unexpected to not have a pack but to also
// to have no new index yet. We set the new index before removing any slots, so
// this should be observable.
return Ok(None);
}
}
}
},
};
let entry = pack.entry(pack_offset);
let res = match pack.decode_header(entry, |id| {
index_file.pack_offset_by_id(id).map(|pack_offset| {
git_pack::data::decode::header::ResolvedBase::InPack(pack.entry(pack_offset))
})
}) {
Ok(header) => Ok(header.into()),
Err(git_pack::data::decode::Error::DeltaBaseUnresolved(base_id)) => {
// Only with multi-pack indices it's allowed to jump to refer to other packs within this
// multi-pack. Otherwise this would constitute a thin pack which is only allowed in transit.
// However, if we somehow end up with that, we will resolve it safely, even though we could
// avoid handling this case and error instead.
let hdr = self
.try_header_inner(
&base_id,
snapshot,
recursion
.map(|r| r.inc_depth())
.or_else(|| DeltaBaseRecursion::new(id).into()),
)
.map_err(|err| Error::DeltaBaseLookup {
err: Box::new(err),
base_id,
id: id.to_owned(),
})?
.ok_or_else(|| Error::DeltaBaseMissing {
base_id,
id: id.to_owned(),
})?;
let handle::index_lookup::Outcome {
object_index:
handle::IndexForObjectInPack {
pack_id: _,
pack_offset,
},
index_file,
pack: possibly_pack,
} = match snapshot.indices[idx].lookup(id) {
Some(res) => res,
None => {
let mut out = None;
for index in snapshot.indices.iter_mut() {
out = index.lookup(id);
if out.is_some() {
break;
}
}
out.unwrap_or_else(|| {
panic!("could not find object {} in any index after looking up one of its base objects {}", id, base_id)
})
}
};
let pack = possibly_pack
.as_ref()
.expect("pack to still be available like just now");
let entry = pack.entry(pack_offset);
pack.decode_header(entry, |id| {
index_file
.pack_offset_by_id(id)
.map(|pack_offset| {
git_pack::data::decode::header::ResolvedBase::InPack(
pack.entry(pack_offset),
)
})
.or_else(|| {
(id == base_id).then(|| {
git_pack::data::decode::header::ResolvedBase::OutOfPack {
kind: hdr.kind(),
num_deltas: hdr.num_deltas(),
}
})
})
})
.map(Into::into)
}
Err(err) => Err(err),
}?;
if idx != 0 {
snapshot.indices.swap(0, idx);
}
return Ok(Some(res));
}
}
}
for lodb in snapshot.loose_dbs.iter() {
// TODO: remove this double-lookup once the borrow checker allows it.
if lodb.contains(id) {
return lodb.try_header(id).map(|opt| opt.map(Into::into)).map_err(Into::into);
}
}
match self.store.load_one_index(self.refresh, snapshot.marker)? {
Some(new_snapshot) => {
*snapshot = new_snapshot;
self.clear_cache();
}
None => return Ok(None),
}
}
}source§impl Store
impl Store
Iteration and traversal
sourcepub fn iter(&self) -> Iter ⓘ
pub fn iter(&self) -> Iter ⓘ
Return an iterator over all objects contained in the database.
The Ids returned by the iterator can typically be used in the locate(…) method.
Note that the result is not sorted or stable, thus ordering can change between runs.
Notes
loose::Iter is used instead of impl Iterator<…> to allow using this iterator in struct fields, as is currently
needed if iterators need to be implemented by hand in the absence of generators.
Examples found in repository?
src/store_impls/dynamic/iter.rs (line 101)
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
pub fn new(db: &dynamic::Store) -> Result<Self, crate::store::load_index::Error> {
let snapshot = db.load_all_indices()?;
let packed_objects = snapshot
.indices
.iter()
.fold(0usize, |dbc, index| dbc.saturating_add(index.num_objects() as usize));
let mut index_iter = snapshot.indices.into_iter();
let loose_dbs = snapshot.loose_dbs;
let order = Default::default();
let state = match index_iter.next() {
Some(index) => {
let num_objects = index.num_objects();
State::Pack {
index_iter,
ordered_entries: maybe_sort_entries(&index, order),
index,
entry_index: 0,
num_objects,
}
}
None => {
let index = 0;
State::Loose {
iter: loose_dbs.get(index).expect("at least one loose db").iter(),
index,
}
}
};
Ok(AllObjects {
state,
loose_dbs,
num_objects: packed_objects,
order,
})
}
}
fn maybe_sort_entries(index: &handle::IndexLookup, order: Ordering) -> Option<Vec<EntryForOrdering>> {
let mut order: Vec<_> = match order {
Ordering::PackLexicographicalThenLooseLexicographical => return None,
Ordering::PackAscendingOffsetThenLooseLexicographical => match &index.file {
// We know that we cannot have more than u32 entry indices per pack.
SingleOrMultiIndex::Single { index, .. } => index
.iter()
.enumerate()
.map(|(idx, e)| EntryForOrdering {
pack_offset: e.pack_offset,
entry_index: idx as u32,
pack_index: 0,
})
.collect(),
SingleOrMultiIndex::Multi { index, .. } => index
.iter()
.enumerate()
.map(|(idx, e)| EntryForOrdering {
pack_offset: e.pack_offset,
entry_index: idx as u32,
pack_index: {
debug_assert!(
e.pack_index < PackId::max_packs_in_multi_index(),
"this shows the relation between u16 and pack_index (u32) and why this is OK"
);
e.pack_index as u16
},
})
.collect(),
},
};
order.sort_by(|a, b| {
a.pack_index
.cmp(&b.pack_index)
.then_with(|| a.pack_offset.cmp(&b.pack_offset))
});
Some(order)
}
impl Iterator for AllObjects {
type Item = Result<ObjectId, loose::iter::Error>;
fn next(&mut self) -> Option<Self::Item> {
match &mut self.state {
State::Depleted => None,
State::Pack {
index_iter,
ordered_entries,
index,
entry_index,
num_objects,
} => {
if *entry_index < *num_objects {
let oid = match ordered_entries {
Some(entries) => index.oid_at_index(entries[*entry_index as usize].entry_index),
None => index.oid_at_index(*entry_index),
}
.to_owned();
*entry_index += 1;
Some(Ok(oid))
} else {
match index_iter.next() {
Some(new_index) => {
*ordered_entries = maybe_sort_entries(&new_index, self.order);
*index = new_index;
*entry_index = 0;
*num_objects = index.num_objects();
}
None => {
let index = 0;
self.state = State::Loose {
iter: self.loose_dbs.get(index).expect("at least one loose odb").iter(),
index,
}
}
}
self.next()
}
}
State::Loose { iter, index } => match iter.next() {
Some(id) => Some(id),
None => {
*index += 1;
match self.loose_dbs.get(*index).map(|ldb| ldb.iter()) {
Some(new_iter) => {
*iter = new_iter;
self.next()
}
None => {
self.state = State::Depleted;
None
}
}
}
},
}
}More examples
src/store_impls/loose/verify.rs (line 57)
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
pub fn verify_integrity(
&self,
mut progress: impl Progress,
should_interrupt: &AtomicBool,
) -> Result<integrity::Statistics, integrity::Error> {
let mut buf = Vec::new();
let sink = crate::sink(self.object_hash);
let mut num_objects = 0;
let start = Instant::now();
let mut progress = progress.add_child_with_id("Validating", *b"VILO"); /* Verify Integrity Loose Objects */
progress.init(None, git_features::progress::count("loose objects"));
for id in self.iter().filter_map(Result::ok) {
let object = self
.try_find(id, &mut buf)
.map_err(|_| integrity::Error::Retry)?
.ok_or(integrity::Error::Retry)?;
let actual_id = sink.write_buf(object.kind, object.data).expect("sink never fails");
if actual_id != id {
return Err(integrity::Error::ObjectHashMismatch {
kind: object.kind,
actual: actual_id,
expected: id,
});
}
object.decode().map_err(|err| integrity::Error::ObjectDecode {
source: err,
kind: object.kind,
id,
})?;
progress.inc();
num_objects += 1;
if should_interrupt.load(Ordering::SeqCst) {
return Err(integrity::Error::Interrupted);
}
}
progress.show_throughput(start);
Ok(integrity::Statistics { num_objects })
}src/store_impls/dynamic/mod.rs (line 155)
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
pub fn structure(&self) -> Result<Vec<Record>, load_index::Error> {
let index = self.index.load();
if !index.is_initialized() {
self.consolidate_with_disk_state(true, false /*load one new index*/)?;
}
let index = self.index.load();
let mut res: Vec<_> = index
.loose_dbs
.iter()
.map(|db| Record::LooseObjectDatabase {
objects_directory: db.path.clone(),
num_objects: db.iter().count(),
})
.collect();
for slot in index.slot_indices.iter().map(|idx| &self.files[*idx]) {
let files = slot.files.load();
let record = match &**files {
Some(index) => {
let state = if index.is_disposable() {
IndexState::Disposable
} else if index.index_is_loaded() {
IndexState::Loaded
} else {
IndexState::Unloaded
};
match index {
IndexAndPacks::Index(b) => Record::Index {
path: b.index.path().into(),
state,
},
IndexAndPacks::MultiIndex(b) => Record::MultiIndex {
path: b.multi_index.path().into(),
state,
},
}
}
None => Record::Empty,
};
res.push(record);
}
Ok(res)
}source§impl Store
impl Store
sourcepub fn verify_integrity(
&self,
progress: impl Progress,
should_interrupt: &AtomicBool
) -> Result<Statistics, Error>
pub fn verify_integrity(
&self,
progress: impl Progress,
should_interrupt: &AtomicBool
) -> Result<Statistics, Error>
Check all loose objects for their integrity checking their hash matches the actual data and by decoding them fully.
Examples found in repository?
src/store_impls/dynamic/verify.rs (lines 218-221)
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
pub fn verify_integrity<C, P, F>(
&self,
mut progress: P,
should_interrupt: &AtomicBool,
options: integrity::Options<F>,
) -> Result<integrity::Outcome<P>, integrity::Error>
where
P: Progress,
C: pack::cache::DecodeEntry,
F: Fn() -> C + Send + Clone,
{
let mut index = self.index.load();
if !index.is_initialized() {
self.consolidate_with_disk_state(true, false)?;
index = self.index.load();
assert!(
index.is_initialized(),
"BUG: after consolidating successfully, we have an initialized index"
)
}
progress.init(
Some(index.slot_indices.len()),
git_features::progress::count("pack indices"),
);
let mut statistics = Vec::new();
let index_check_message = |path: &std::path::Path| {
format!(
"Checking integrity: {}",
path.file_name()
.map(|f| f.to_string_lossy())
.unwrap_or_else(std::borrow::Cow::default)
)
};
for slot_index in &index.slot_indices {
let slot = &self.files[*slot_index];
if slot.generation.load(Ordering::SeqCst) != index.generation {
return Err(integrity::Error::NeedsRetryDueToChangeOnDisk);
}
let files = slot.files.load();
let files = Option::as_ref(&files).ok_or(integrity::Error::NeedsRetryDueToChangeOnDisk)?;
let start = Instant::now();
let (mut child_progress, num_objects, index_path) = match files {
IndexAndPacks::Index(bundle) => {
let index;
let index = match bundle.index.loaded() {
Some(index) => index.deref(),
None => {
index = pack::index::File::at(bundle.index.path(), self.object_hash)?;
&index
}
};
let pack;
let data = match bundle.data.loaded() {
Some(pack) => pack.deref(),
None => {
pack = pack::data::File::at(bundle.data.path(), self.object_hash)?;
&pack
}
};
let outcome = index.verify_integrity(
Some(pack::index::verify::PackContext {
data,
options: options.clone(),
}),
progress.add_child_with_id("never shown", git_features::progress::UNKNOWN),
should_interrupt,
)?;
statistics.push(IndexStatistics {
path: bundle.index.path().to_owned(),
statistics: SingleOrMultiStatistics::Single(
outcome
.pack_traverse_statistics
.expect("pack provided so there are stats"),
),
});
(outcome.progress, index.num_objects(), index.path().to_owned())
}
IndexAndPacks::MultiIndex(bundle) => {
let index;
let index = match bundle.multi_index.loaded() {
Some(index) => index.deref(),
None => {
index = pack::multi_index::File::at(bundle.multi_index.path())?;
&index
}
};
let outcome = index.verify_integrity(
progress.add_child_with_id("never shown", git_features::progress::UNKNOWN),
should_interrupt,
options.clone(),
)?;
let index_dir = bundle.multi_index.path().parent().expect("file in a directory");
statistics.push(IndexStatistics {
path: Default::default(),
statistics: SingleOrMultiStatistics::Multi(
outcome
.pack_traverse_statistics
.into_iter()
.zip(index.index_names())
.map(|(statistics, index_name)| (index_dir.join(index_name), statistics))
.collect(),
),
});
(outcome.progress, index.num_objects(), index.path().to_owned())
}
};
child_progress.set_name(index_check_message(&index_path));
child_progress.show_throughput_with(
start,
num_objects as usize,
git_features::progress::count("objects").expect("set"),
MessageLevel::Success,
);
progress.inc();
}
progress.init(
Some(index.loose_dbs.len()),
git_features::progress::count("loose object stores"),
);
let mut loose_object_stores = Vec::new();
for loose_db in &*index.loose_dbs {
let out = loose_db
.verify_integrity(
progress.add_child_with_id(loose_db.path().display().to_string(), *b"VISP"), /* Verify Integrity Store Path */
should_interrupt,
)
.map(|statistics| integrity::LooseObjectStatistics {
path: loose_db.path().to_owned(),
statistics,
})?;
loose_object_stores.push(out);
}
Ok(integrity::Outcome {
loose_object_stores,
index_statistics: statistics,
progress,
})
}source§impl Store
impl Store
Initialization
sourcepub fn at(objects_directory: impl Into<PathBuf>, object_hash: Kind) -> Store
pub fn at(objects_directory: impl Into<PathBuf>, object_hash: Kind) -> Store
Initialize the Db with the objects_directory containing the hexadecimal first byte subdirectories, which in turn
contain all loose objects.
In a git repository, this would be .git/objects.
The object_hash determines which hash to use when writing, finding or iterating objects.
Examples found in repository?
src/store_impls/dynamic/load_index.rs (line 222)
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419
pub(crate) fn consolidate_with_disk_state(
&self,
needs_init: bool,
load_new_index: bool,
) -> Result<Option<Snapshot>, Error> {
let index = self.index.load();
let previous_index_state = Arc::as_ptr(&index) as usize;
// IMPORTANT: get a lock after we recorded the previous state.
let write = self.write.lock();
let objects_directory = &self.path;
// Now we know the index isn't going to change anymore, even though threads might still load indices in the meantime.
let index = self.index.load();
if previous_index_state != Arc::as_ptr(&index) as usize {
// Someone else took the look before and changed the index. Return it without doing any additional work.
return Ok(Some(self.collect_snapshot()));
}
let was_uninitialized = !index.is_initialized();
// We might not be able to detect by pointer if the state changed, as this itself is racy. So we keep track of double-initialization
// using a flag, which means that if `needs_init` was true we saw the index uninitialized once, but now that we are here it's
// initialized meaning that somebody was faster and we couldn't detect it by comparisons to the index.
// If so, make sure we collect the snapshot instead of returning None in case nothing actually changed, which is likely with a
// race like this.
if !was_uninitialized && needs_init {
return Ok(Some(self.collect_snapshot()));
}
self.num_disk_state_consolidation.fetch_add(1, Ordering::Relaxed);
let db_paths: Vec<_> = std::iter::once(objects_directory.to_owned())
.chain(crate::alternate::resolve(objects_directory, &self.current_dir)?)
.collect();
// turn db paths into loose object databases. Reuse what's there, but only if it is in the right order.
let loose_dbs = if was_uninitialized
|| db_paths.len() != index.loose_dbs.len()
|| db_paths
.iter()
.zip(index.loose_dbs.iter().map(|ldb| &ldb.path))
.any(|(lhs, rhs)| lhs != rhs)
{
Arc::new(
db_paths
.iter()
.map(|path| crate::loose::Store::at(path, self.object_hash))
.collect::<Vec<_>>(),
)
} else {
Arc::clone(&index.loose_dbs)
};
let indices_by_modification_time = Self::collect_indices_and_mtime_sorted_by_size(
db_paths,
index.slot_indices.len().into(),
self.use_multi_pack_index.then(|| self.object_hash),
)?;
let mut idx_by_index_path: BTreeMap<_, _> = index
.slot_indices
.iter()
.filter_map(|&idx| {
let f = &self.files[idx];
Option::as_ref(&f.files.load()).map(|f| (f.index_path().to_owned(), idx))
})
.collect();
let mut new_slot_map_indices = Vec::new(); // these indices into the slot map still exist there/didn't change
let mut index_paths_to_add = was_uninitialized
.then(|| VecDeque::with_capacity(indices_by_modification_time.len()))
.unwrap_or_default();
// Figure out this number based on what we see while handling the existing indices
let mut num_loaded_indices = 0;
for (index_info, mtime) in indices_by_modification_time.into_iter().map(|(a, b, _)| (a, b)) {
match idx_by_index_path.remove(index_info.path()) {
Some(slot_idx) => {
let slot = &self.files[slot_idx];
let files_guard = slot.files.load();
let files =
Option::as_ref(&files_guard).expect("slot is set or we wouldn't know it points to this file");
if index_info.is_multi_index() && files.mtime() != mtime {
// we have a changed multi-pack index. We can't just change the existing slot as it may alter slot indices
// that are currently available. Instead we have to move what's there into a new slot, along with the changes,
// and later free the slot or dispose of the index in the slot (like we do for removed/missing files).
index_paths_to_add.push_back((index_info, mtime, Some(slot_idx)));
// If the current slot is loaded, the soon-to-be copied multi-index path will be loaded as well.
if files.index_is_loaded() {
num_loaded_indices += 1;
}
} else {
// packs and indices are immutable, so no need to check modification times. Unchanged multi-pack indices also
// are handled like this just to be sure they are in the desired state. For these, the only way this could happen
// is if somebody deletes and then puts back
if Self::assure_slot_matches_index(&write, slot, index_info, mtime, index.generation) {
num_loaded_indices += 1;
}
new_slot_map_indices.push(slot_idx);
}
}
None => index_paths_to_add.push_back((index_info, mtime, None)),
}
}
let needs_stable_indices = self.maintain_stable_indices(&write);
let mut next_possibly_free_index = index
.slot_indices
.iter()
.max()
.map(|idx| (idx + 1) % self.files.len())
.unwrap_or(0);
let mut num_indices_checked = 0;
let mut needs_generation_change = false;
let mut slot_indices_to_remove: Vec<_> = idx_by_index_path.into_values().collect();
while let Some((mut index_info, mtime, move_from_slot_idx)) = index_paths_to_add.pop_front() {
'increment_slot_index: loop {
if num_indices_checked == self.files.len() {
return Err(Error::InsufficientSlots {
current: self.files.len(),
needed: index_paths_to_add.len() + 1, /*the one currently popped off*/
});
}
let slot_index = next_possibly_free_index;
let slot = &self.files[slot_index];
next_possibly_free_index = (next_possibly_free_index + 1) % self.files.len();
num_indices_checked += 1;
match move_from_slot_idx {
Some(move_from_slot_idx) => {
debug_assert!(index_info.is_multi_index(), "only set for multi-pack indices");
if slot_index == move_from_slot_idx {
// don't try to move onto ourselves
continue 'increment_slot_index;
}
match Self::try_set_index_slot(
&write,
slot,
index_info,
mtime,
index.generation,
needs_stable_indices,
) {
Ok(dest_was_empty) => {
slot_indices_to_remove.push(move_from_slot_idx);
new_slot_map_indices.push(slot_index);
// To avoid handling out the wrong pack (due to reassigned pack ids), declare this a new generation.
if !dest_was_empty {
needs_generation_change = true;
}
break 'increment_slot_index;
}
Err(unused_index_info) => index_info = unused_index_info,
}
}
None => {
match Self::try_set_index_slot(
&write,
slot,
index_info,
mtime,
index.generation,
needs_stable_indices,
) {
Ok(dest_was_empty) => {
new_slot_map_indices.push(slot_index);
if !dest_was_empty {
needs_generation_change = true;
}
break 'increment_slot_index;
}
Err(unused_index_info) => index_info = unused_index_info,
}
}
}
// This isn't racy as it's only us who can change the Option::Some/None state of a slot.
}
}
assert_eq!(
index_paths_to_add.len(),
0,
"By this time we have assigned all new files to slots"
);
let generation = if needs_generation_change {
index.generation.checked_add(1).ok_or(Error::GenerationOverflow)?
} else {
index.generation
};
let index_unchanged = index.slot_indices == new_slot_map_indices;
if generation != index.generation {
assert!(
!index_unchanged,
"if the generation changed, the slot index must have changed for sure"
);
}
if !index_unchanged || loose_dbs != index.loose_dbs {
let new_index = Arc::new(SlotMapIndex {
slot_indices: new_slot_map_indices,
loose_dbs,
generation,
// if there was a prior generation, some indices might already be loaded. But we deal with it by trying to load the next index then,
// until we find one.
next_index_to_load: index_unchanged
.then(|| Arc::clone(&index.next_index_to_load))
.unwrap_or_default(),
loaded_indices: index_unchanged
.then(|| Arc::clone(&index.loaded_indices))
.unwrap_or_else(|| Arc::new(num_loaded_indices.into())),
num_indices_currently_being_loaded: Default::default(),
});
self.index.store(new_index);
}
// deleted items - remove their slots AFTER we have set the new index if we may alter indices, otherwise we only declare them garbage.
// removing slots may cause pack loading to fail, and they will then reload their indices.
for slot in slot_indices_to_remove.into_iter().map(|idx| &self.files[idx]) {
let _lock = slot.write.lock();
let mut files = slot.files.load_full();
let files_mut = Arc::make_mut(&mut files);
if needs_stable_indices {
if let Some(files) = files_mut.as_mut() {
files.trash();
// generation stays the same, as it's the same value still but scheduled for eventual removal.
}
} else {
*files_mut = None;
};
slot.files.store(files);
if !needs_stable_indices {
// Not racy due to lock, generation must be set after unsetting the slot value AND storing it.
slot.generation.store(generation, Ordering::SeqCst);
}
}
let new_index = self.index.load();
Ok(if index.state_id() == new_index.state_id() {
// there was no change, and nothing was loaded in the meantime, reflect that in the return value to not get into loops
None
} else {
if load_new_index {
self.load_next_index(new_index);
}
Some(self.collect_snapshot())
})
}sourcepub fn path(&self) -> &Path
pub fn path(&self) -> &Path
Return the path to our objects directory.
Examples found in repository?
src/store_impls/dynamic/verify.rs (line 219)
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
pub fn verify_integrity<C, P, F>(
&self,
mut progress: P,
should_interrupt: &AtomicBool,
options: integrity::Options<F>,
) -> Result<integrity::Outcome<P>, integrity::Error>
where
P: Progress,
C: pack::cache::DecodeEntry,
F: Fn() -> C + Send + Clone,
{
let mut index = self.index.load();
if !index.is_initialized() {
self.consolidate_with_disk_state(true, false)?;
index = self.index.load();
assert!(
index.is_initialized(),
"BUG: after consolidating successfully, we have an initialized index"
)
}
progress.init(
Some(index.slot_indices.len()),
git_features::progress::count("pack indices"),
);
let mut statistics = Vec::new();
let index_check_message = |path: &std::path::Path| {
format!(
"Checking integrity: {}",
path.file_name()
.map(|f| f.to_string_lossy())
.unwrap_or_else(std::borrow::Cow::default)
)
};
for slot_index in &index.slot_indices {
let slot = &self.files[*slot_index];
if slot.generation.load(Ordering::SeqCst) != index.generation {
return Err(integrity::Error::NeedsRetryDueToChangeOnDisk);
}
let files = slot.files.load();
let files = Option::as_ref(&files).ok_or(integrity::Error::NeedsRetryDueToChangeOnDisk)?;
let start = Instant::now();
let (mut child_progress, num_objects, index_path) = match files {
IndexAndPacks::Index(bundle) => {
let index;
let index = match bundle.index.loaded() {
Some(index) => index.deref(),
None => {
index = pack::index::File::at(bundle.index.path(), self.object_hash)?;
&index
}
};
let pack;
let data = match bundle.data.loaded() {
Some(pack) => pack.deref(),
None => {
pack = pack::data::File::at(bundle.data.path(), self.object_hash)?;
&pack
}
};
let outcome = index.verify_integrity(
Some(pack::index::verify::PackContext {
data,
options: options.clone(),
}),
progress.add_child_with_id("never shown", git_features::progress::UNKNOWN),
should_interrupt,
)?;
statistics.push(IndexStatistics {
path: bundle.index.path().to_owned(),
statistics: SingleOrMultiStatistics::Single(
outcome
.pack_traverse_statistics
.expect("pack provided so there are stats"),
),
});
(outcome.progress, index.num_objects(), index.path().to_owned())
}
IndexAndPacks::MultiIndex(bundle) => {
let index;
let index = match bundle.multi_index.loaded() {
Some(index) => index.deref(),
None => {
index = pack::multi_index::File::at(bundle.multi_index.path())?;
&index
}
};
let outcome = index.verify_integrity(
progress.add_child_with_id("never shown", git_features::progress::UNKNOWN),
should_interrupt,
options.clone(),
)?;
let index_dir = bundle.multi_index.path().parent().expect("file in a directory");
statistics.push(IndexStatistics {
path: Default::default(),
statistics: SingleOrMultiStatistics::Multi(
outcome
.pack_traverse_statistics
.into_iter()
.zip(index.index_names())
.map(|(statistics, index_name)| (index_dir.join(index_name), statistics))
.collect(),
),
});
(outcome.progress, index.num_objects(), index.path().to_owned())
}
};
child_progress.set_name(index_check_message(&index_path));
child_progress.show_throughput_with(
start,
num_objects as usize,
git_features::progress::count("objects").expect("set"),
MessageLevel::Success,
);
progress.inc();
}
progress.init(
Some(index.loose_dbs.len()),
git_features::progress::count("loose object stores"),
);
let mut loose_object_stores = Vec::new();
for loose_db in &*index.loose_dbs {
let out = loose_db
.verify_integrity(
progress.add_child_with_id(loose_db.path().display().to_string(), *b"VISP"), /* Verify Integrity Store Path */
should_interrupt,
)
.map(|statistics| integrity::LooseObjectStatistics {
path: loose_db.path().to_owned(),
statistics,
})?;
loose_object_stores.push(out);
}
Ok(integrity::Outcome {
loose_object_stores,
index_statistics: statistics,
progress,
})
}sourcepub fn object_hash(&self) -> Kind
pub fn object_hash(&self) -> Kind
Return the kind of hash we would iterate and write.
Trait Implementations§
source§impl PartialEq<Store> for Store
impl PartialEq<Store> for Store
source§impl Write for Store
impl Write for Store
source§fn write_buf(&self, kind: Kind, from: &[u8]) -> Result<ObjectId, Self::Error>
fn write_buf(&self, kind: Kind, from: &[u8]) -> Result<ObjectId, Self::Error>
Write the given buffer in from to disk in one syscall at best.
This will cost at least 4 IO operations.
impl Eq for Store
impl StructuralEq for Store
impl StructuralPartialEq for Store
Auto Trait Implementations§
impl RefUnwindSafe for Store
impl Send for Store
impl Sync for Store
impl Unpin for Store
impl UnwindSafe for Store
Blanket Implementations§
§impl<Q, K> Equivalent<K> for Qwhere
Q: Eq + ?Sized,
K: Borrow<Q> + ?Sized,
impl<Q, K> Equivalent<K> for Qwhere
Q: Eq + ?Sized,
K: Borrow<Q> + ?Sized,
§fn equivalent(&self, key: &K) -> bool
fn equivalent(&self, key: &K) -> bool
Checks if this value is equivalent to the given key. Read more