nix-super/src/libexpr/eval-cache.cc

767 lines
23 KiB
C++
Raw Normal View History

#include "eval-cache.hh"
#include "sqlite.hh"
2019-06-07 23:38:39 +03:00
#include "eval.hh"
2020-04-20 14:13:52 +03:00
#include "eval-inline.hh"
#include "store-api.hh"
2020-04-20 14:13:52 +03:00
namespace nix::eval_cache {
static const char * schema = R"sql(
create table if not exists Attributes (
2020-04-20 14:13:52 +03:00
parent integer not null,
name text,
type integer not null,
value text,
2020-06-29 20:08:37 +03:00
context text,
2020-04-20 14:13:52 +03:00
primary key (parent, name)
);
)sql";
2020-04-20 14:13:52 +03:00
struct AttrDb
{
std::atomic_bool failed{false};
const Store & cfg;
2020-04-20 14:13:52 +03:00
struct State
{
SQLite db;
SQLiteStmt insertAttribute;
2020-06-29 20:08:37 +03:00
SQLiteStmt insertAttributeWithContext;
2020-04-20 14:13:52 +03:00
SQLiteStmt queryAttribute;
SQLiteStmt queryAttributes;
std::unique_ptr<SQLiteTxn> txn;
};
std::unique_ptr<Sync<State>> _state;
SymbolTable & symbols;
AttrDb(
const Store & cfg,
const Hash & fingerprint,
SymbolTable & symbols)
: cfg(cfg)
, _state(std::make_unique<Sync<State>>())
, symbols(symbols)
2020-04-20 14:13:52 +03:00
{
auto state(_state->lock());
Use `std::set<StringContextElem>` not `PathSet` for string contexts Motivation `PathSet` is not correct because string contexts have other forms (`Built` and `DrvDeep`) that are not rendered as plain store paths. Instead of wrongly using `PathSet`, or "stringly typed" using `StringSet`, use `std::std<StringContextElem>`. ----- In support of this change, `NixStringContext` is now defined as `std::std<StringContextElem>` not `std:vector<StringContextElem>`. The old definition was just used by a `getContext` method which was only used by the eval cache. It can be deleted altogether since the types are now unified and the preexisting `copyContext` function already suffices. Summarizing the previous paragraph: Old: - `value/context.hh`: `NixStringContext = std::vector<StringContextElem>` - `value.hh`: `NixStringContext Value::getContext(...)` - `value.hh`: `copyContext(...)` New: - `value/context.hh`: `NixStringContext = std::set<StringContextElem>` - `value.hh`: `copyContext(...)` ---- The string representation of string context elements no longer contains the store dir. The diff of `src/libexpr/tests/value/context.cc` should make clear what the new representation is, so we recommend reviewing that file first. This was done for two reasons: Less API churn: `Value::mkString` and friends did not take a `Store` before. But if `NixStringContextElem::{parse, to_string}` *do* take a store (as they did before), then we cannot have the `Value` functions use them (in order to work with the fully-structured `NixStringContext`) without adding that argument. That would have been a lot of churn of threading the store, and this diff is already large enough, so the easier and less invasive thing to do was simply make the element `parse` and `to_string` functions not take the `Store` reference, and the easiest way to do that was to simply drop the store dir. Space usage: Dropping the `/nix/store/` (or similar) from the internal representation will safe space in the heap of the Nix programming being interpreted. If the heap contains many strings with non-trivial contexts, the saving could add up to something significant. ---- The eval cache version is bumped. The eval cache serialization uses `NixStringContextElem::{parse, to_string}`, and since those functions are changed per the above, that means the on-disk representation is also changed. This is simply done by changing the name of the used for the eval cache from `eval-cache-v4` to eval-cache-v5`. ---- To avoid some duplication `EvalCache::mkPathString` is added to abstract over the simple case of turning a store path to a string with just that string in the context. Context This PR picks up where #7543 left off. That one introduced the fully structured `NixStringContextElem` data type, but kept `PathSet context` as an awkward middle ground between internal `char[][]` interpreter heap string contexts and `NixStringContext` fully parsed string contexts. The infelicity of `PathSet context` was specifically called out during Nix team group review, but it was agreeing that fixing it could be left as future work. This is that future work. A possible follow-up step would be to get rid of the `char[][]` evaluator heap representation, too, but it is not yet clear how to do that. To use `NixStringContextElem` there we would need to get the STL containers to GC pointers in the GC build, and I am not sure how to do that. ---- PR #7543 effectively is writing the inverse of a `mkPathString`, `mkOutputString`, and one more such function for the `DrvDeep` case. I would like that PR to have property tests ensuring it is actually the inverse as expected. This PR sets things up nicely so that reworking that PR to be in that more elegant and better tested way is possible. Co-authored-by: Théophane Hufschmitt <7226587+thufschmitt@users.noreply.github.com>
2023-01-29 03:31:10 +02:00
Path cacheDir = getCacheDir() + "/nix/eval-cache-v5";
2020-04-20 14:13:52 +03:00
createDirs(cacheDir);
Path dbPath = cacheDir + "/" + fingerprint.to_string(HashFormat::Base16, false) + ".sqlite";
2020-04-20 14:13:52 +03:00
state->db = SQLite(dbPath);
state->db.isCache();
state->db.exec(schema);
state->insertAttribute.create(state->db,
"insert or replace into Attributes(parent, name, type, value) values (?, ?, ?, ?)");
2020-06-29 20:08:37 +03:00
state->insertAttributeWithContext.create(state->db,
"insert or replace into Attributes(parent, name, type, value, context) values (?, ?, ?, ?, ?)");
2020-04-20 14:13:52 +03:00
state->queryAttribute.create(state->db,
2020-06-29 20:08:37 +03:00
"select rowid, type, value, context from Attributes where parent = ? and name = ?");
2020-04-20 14:13:52 +03:00
state->queryAttributes.create(state->db,
"select name from Attributes where parent = ?");
state->txn = std::make_unique<SQLiteTxn>(state->db);
}
~AttrDb()
{
try {
auto state(_state->lock());
if (!failed)
state->txn->commit();
2020-04-20 14:13:52 +03:00
state->txn.reset();
} catch (...) {
ignoreException();
}
}
template<typename F>
AttrId doSQLite(F && fun)
{
if (failed) return 0;
try {
return fun();
} catch (SQLiteError &) {
ignoreException();
failed = true;
return 0;
}
}
2020-04-20 14:13:52 +03:00
AttrId setAttrs(
AttrKey key,
const std::vector<Symbol> & attrs)
{
return doSQLite([&]()
{
auto state(_state->lock());
2020-04-20 14:13:52 +03:00
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::FullAttrs)
2020-04-20 14:13:52 +03:00
(0, false).exec();
AttrId rowId = state->db.getLastInsertedRowId();
assert(rowId);
for (auto & attr : attrs)
state->insertAttribute.use()
(rowId)
(symbols[attr])
(AttrType::Placeholder)
(0, false).exec();
return rowId;
});
2020-04-20 14:13:52 +03:00
}
AttrId setString(
AttrKey key,
2020-06-29 20:08:37 +03:00
std::string_view s,
const char * * context = nullptr)
2020-04-20 14:13:52 +03:00
{
return doSQLite([&]()
{
auto state(_state->lock());
2020-04-20 14:13:52 +03:00
if (context) {
std::string ctx;
for (const char * * p = context; *p; ++p) {
if (p != context) ctx.push_back(' ');
ctx.append(*p);
}
state->insertAttributeWithContext.use()
(key.first)
(symbols[key.second])
(AttrType::String)
(s)
(ctx).exec();
} else {
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::String)
2020-06-29 20:08:37 +03:00
(s).exec();
}
2020-04-20 14:13:52 +03:00
return state->db.getLastInsertedRowId();
});
2020-04-20 14:13:52 +03:00
}
AttrId setBool(
AttrKey key,
bool b)
{
return doSQLite([&]()
{
auto state(_state->lock());
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::Bool)
(b ? 1 : 0).exec();
return state->db.getLastInsertedRowId();
});
}
AttrId setInt(
AttrKey key,
int n)
{
return doSQLite([&]()
{
auto state(_state->lock());
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::Int)
(n).exec();
return state->db.getLastInsertedRowId();
});
}
AttrId setListOfStrings(
AttrKey key,
const std::vector<std::string> & l)
{
return doSQLite([&]()
{
auto state(_state->lock());
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::ListOfStrings)
(concatStringsSep("\t", l)).exec();
return state->db.getLastInsertedRowId();
});
}
2020-04-20 14:13:52 +03:00
AttrId setPlaceholder(AttrKey key)
{
return doSQLite([&]()
{
auto state(_state->lock());
2020-04-20 14:13:52 +03:00
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::Placeholder)
(0, false).exec();
2020-04-20 14:13:52 +03:00
return state->db.getLastInsertedRowId();
});
2020-04-20 14:13:52 +03:00
}
AttrId setMissing(AttrKey key)
{
return doSQLite([&]()
{
auto state(_state->lock());
2020-04-20 14:13:52 +03:00
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::Missing)
(0, false).exec();
2020-04-20 14:13:52 +03:00
return state->db.getLastInsertedRowId();
});
2020-04-20 14:13:52 +03:00
}
AttrId setMisc(AttrKey key)
{
return doSQLite([&]()
{
auto state(_state->lock());
2020-04-20 14:13:52 +03:00
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::Misc)
(0, false).exec();
2020-04-20 14:13:52 +03:00
return state->db.getLastInsertedRowId();
});
2020-04-20 14:13:52 +03:00
}
AttrId setFailed(AttrKey key)
{
return doSQLite([&]()
{
auto state(_state->lock());
2020-04-20 14:13:52 +03:00
state->insertAttribute.use()
(key.first)
(symbols[key.second])
(AttrType::Failed)
(0, false).exec();
2020-04-20 14:13:52 +03:00
return state->db.getLastInsertedRowId();
});
2020-04-20 14:13:52 +03:00
}
std::optional<std::pair<AttrId, AttrValue>> getAttr(AttrKey key)
2020-04-20 14:13:52 +03:00
{
auto state(_state->lock());
auto queryAttribute(state->queryAttribute.use()(key.first)(symbols[key.second]));
2020-04-20 14:13:52 +03:00
if (!queryAttribute.next()) return {};
auto rowId = (AttrId) queryAttribute.getInt(0);
2020-04-20 14:13:52 +03:00
auto type = (AttrType) queryAttribute.getInt(1);
switch (type) {
case AttrType::Placeholder:
return {{rowId, placeholder_t()}};
case AttrType::FullAttrs: {
// FIXME: expensive, should separate this out.
std::vector<Symbol> attrs;
auto queryAttributes(state->queryAttributes.use()(rowId));
while (queryAttributes.next())
attrs.emplace_back(symbols.create(queryAttributes.getStr(0)));
return {{rowId, attrs}};
}
2020-06-29 20:08:37 +03:00
case AttrType::String: {
NixStringContext context;
2020-06-29 20:08:37 +03:00
if (!queryAttribute.isNull(3))
for (auto & s : tokenizeString<std::vector<std::string>>(queryAttribute.getStr(3), ";"))
Use `std::set<StringContextElem>` not `PathSet` for string contexts Motivation `PathSet` is not correct because string contexts have other forms (`Built` and `DrvDeep`) that are not rendered as plain store paths. Instead of wrongly using `PathSet`, or "stringly typed" using `StringSet`, use `std::std<StringContextElem>`. ----- In support of this change, `NixStringContext` is now defined as `std::std<StringContextElem>` not `std:vector<StringContextElem>`. The old definition was just used by a `getContext` method which was only used by the eval cache. It can be deleted altogether since the types are now unified and the preexisting `copyContext` function already suffices. Summarizing the previous paragraph: Old: - `value/context.hh`: `NixStringContext = std::vector<StringContextElem>` - `value.hh`: `NixStringContext Value::getContext(...)` - `value.hh`: `copyContext(...)` New: - `value/context.hh`: `NixStringContext = std::set<StringContextElem>` - `value.hh`: `copyContext(...)` ---- The string representation of string context elements no longer contains the store dir. The diff of `src/libexpr/tests/value/context.cc` should make clear what the new representation is, so we recommend reviewing that file first. This was done for two reasons: Less API churn: `Value::mkString` and friends did not take a `Store` before. But if `NixStringContextElem::{parse, to_string}` *do* take a store (as they did before), then we cannot have the `Value` functions use them (in order to work with the fully-structured `NixStringContext`) without adding that argument. That would have been a lot of churn of threading the store, and this diff is already large enough, so the easier and less invasive thing to do was simply make the element `parse` and `to_string` functions not take the `Store` reference, and the easiest way to do that was to simply drop the store dir. Space usage: Dropping the `/nix/store/` (or similar) from the internal representation will safe space in the heap of the Nix programming being interpreted. If the heap contains many strings with non-trivial contexts, the saving could add up to something significant. ---- The eval cache version is bumped. The eval cache serialization uses `NixStringContextElem::{parse, to_string}`, and since those functions are changed per the above, that means the on-disk representation is also changed. This is simply done by changing the name of the used for the eval cache from `eval-cache-v4` to eval-cache-v5`. ---- To avoid some duplication `EvalCache::mkPathString` is added to abstract over the simple case of turning a store path to a string with just that string in the context. Context This PR picks up where #7543 left off. That one introduced the fully structured `NixStringContextElem` data type, but kept `PathSet context` as an awkward middle ground between internal `char[][]` interpreter heap string contexts and `NixStringContext` fully parsed string contexts. The infelicity of `PathSet context` was specifically called out during Nix team group review, but it was agreeing that fixing it could be left as future work. This is that future work. A possible follow-up step would be to get rid of the `char[][]` evaluator heap representation, too, but it is not yet clear how to do that. To use `NixStringContextElem` there we would need to get the STL containers to GC pointers in the GC build, and I am not sure how to do that. ---- PR #7543 effectively is writing the inverse of a `mkPathString`, `mkOutputString`, and one more such function for the `DrvDeep` case. I would like that PR to have property tests ensuring it is actually the inverse as expected. This PR sets things up nicely so that reworking that PR to be in that more elegant and better tested way is possible. Co-authored-by: Théophane Hufschmitt <7226587+thufschmitt@users.noreply.github.com>
2023-01-29 03:31:10 +02:00
context.insert(NixStringContextElem::parse(s));
2020-06-29 20:08:37 +03:00
return {{rowId, string_t{queryAttribute.getStr(2), context}}};
}
case AttrType::Bool:
return {{rowId, queryAttribute.getInt(2) != 0}};
case AttrType::Int:
return {{rowId, int_t{queryAttribute.getInt(2)}}};
case AttrType::ListOfStrings:
return {{rowId, tokenizeString<std::vector<std::string>>(queryAttribute.getStr(2), "\t")}};
case AttrType::Missing:
return {{rowId, missing_t()}};
case AttrType::Misc:
return {{rowId, misc_t()}};
case AttrType::Failed:
return {{rowId, failed_t()}};
default:
throw Error("unexpected type in evaluation cache");
}
2020-04-20 14:13:52 +03:00
}
};
static std::shared_ptr<AttrDb> makeAttrDb(
const Store & cfg,
const Hash & fingerprint,
SymbolTable & symbols)
{
try {
return std::make_shared<AttrDb>(cfg, fingerprint, symbols);
} catch (SQLiteError &) {
ignoreException();
return nullptr;
}
}
2020-04-20 14:13:52 +03:00
EvalCache::EvalCache(
std::optional<std::reference_wrapper<const Hash>> useCache,
2020-04-20 14:13:52 +03:00
EvalState & state,
RootLoader rootLoader)
: db(useCache ? makeAttrDb(*state.store, *useCache, state.symbols) : nullptr)
2020-04-20 14:13:52 +03:00
, state(state)
, rootLoader(rootLoader)
{
}
Value * EvalCache::getRootValue()
{
if (!value) {
debug("getting root value");
value = allocRootValue(rootLoader());
}
return *value;
}
ref<AttrCursor> EvalCache::getRoot()
{
return make_ref<AttrCursor>(ref(shared_from_this()), std::nullopt);
2020-04-20 14:13:52 +03:00
}
2020-04-20 14:13:52 +03:00
AttrCursor::AttrCursor(
ref<EvalCache> root,
Parent parent,
Value * value,
std::optional<std::pair<AttrId, AttrValue>> && cachedValue)
: root(root), parent(parent), cachedValue(std::move(cachedValue))
{
if (value)
_value = allocRootValue(value);
}
2020-04-20 14:13:52 +03:00
AttrKey AttrCursor::getKey()
{
if (!parent)
return {0, root->state.sEpsilon};
if (!parent->first->cachedValue) {
parent->first->cachedValue = root->db->getAttr(parent->first->getKey());
2020-04-20 14:13:52 +03:00
assert(parent->first->cachedValue);
}
return {parent->first->cachedValue->first, parent->second};
}
2020-04-20 14:13:52 +03:00
Value & AttrCursor::getValue()
{
if (!_value) {
if (parent) {
auto & vParent = parent->first->getValue();
root->state.forceAttrs(vParent, noPos, "while searching for an attribute");
2020-04-20 14:13:52 +03:00
auto attr = vParent.attrs->get(parent->second);
if (!attr)
throw Error("attribute '%s' is unexpectedly missing", getAttrPathStr());
_value = allocRootValue(attr->value);
} else
_value = allocRootValue(root->getRootValue());
}
return **_value;
}
std::vector<Symbol> AttrCursor::getAttrPath() const
{
if (parent) {
auto attrPath = parent->first->getAttrPath();
attrPath.push_back(parent->second);
return attrPath;
} else
return {};
}
2020-04-20 14:13:52 +03:00
std::vector<Symbol> AttrCursor::getAttrPath(Symbol name) const
{
auto attrPath = getAttrPath();
attrPath.push_back(name);
return attrPath;
}
2020-04-20 14:13:52 +03:00
std::string AttrCursor::getAttrPathStr() const
{
return concatStringsSep(".", root->state.symbols.resolve(getAttrPath()));
}
2020-04-20 14:13:52 +03:00
std::string AttrCursor::getAttrPathStr(Symbol name) const
{
return concatStringsSep(".", root->state.symbols.resolve(getAttrPath(name)));
2020-04-20 14:13:52 +03:00
}
2020-04-20 14:13:52 +03:00
Value & AttrCursor::forceValue()
{
debug("evaluating uncached attribute '%s'", getAttrPathStr());
2019-06-07 23:38:39 +03:00
2020-04-20 14:13:52 +03:00
auto & v = getValue();
2020-04-20 14:13:52 +03:00
try {
root->state.forceValue(v, noPos);
2020-04-20 14:13:52 +03:00
} catch (EvalError &) {
debug("setting '%s' to failed", getAttrPathStr());
if (root->db)
cachedValue = {root->db->setFailed(getKey()), failed_t()};
throw;
}
2020-04-20 14:13:52 +03:00
if (root->db && (!cachedValue || std::get_if<placeholder_t>(&cachedValue->second))) {
if (v.type() == nString)
cachedValue = {root->db->setString(getKey(), v.c_str(), v.context()),
string_t{v.c_str(), {}}};
else if (v.type() == nPath) {
auto path = v.path().path;
cachedValue = {root->db->setString(getKey(), path.abs()), string_t{path.abs(), {}}};
}
2020-12-29 03:40:04 +02:00
else if (v.type() == nBool)
cachedValue = {root->db->setBool(getKey(), v.boolean), v.boolean};
else if (v.type() == nInt)
cachedValue = {root->db->setInt(getKey(), v.integer), int_t{v.integer}};
else if (v.type() == nAttrs)
2020-04-20 14:13:52 +03:00
; // FIXME: do something?
else
cachedValue = {root->db->setMisc(getKey()), misc_t()};
}
return v;
}
Suggestions AttrCursor::getSuggestionsForAttr(Symbol name)
{
auto attrNames = getAttrs();
std::set<std::string> strAttrNames;
for (auto & name : attrNames)
strAttrNames.insert(root->state.symbols[name]);
return Suggestions::bestMatches(strAttrNames, root->state.symbols[name]);
}
std::shared_ptr<AttrCursor> AttrCursor::maybeGetAttr(Symbol name, bool forceErrors)
{
2020-04-20 14:13:52 +03:00
if (root->db) {
if (!cachedValue)
cachedValue = root->db->getAttr(getKey());
2020-04-20 14:13:52 +03:00
if (cachedValue) {
if (auto attrs = std::get_if<std::vector<Symbol>>(&cachedValue->second)) {
for (auto & attr : *attrs)
if (attr == name)
return std::make_shared<AttrCursor>(root, std::make_pair(shared_from_this(), attr));
2020-04-20 14:13:52 +03:00
return nullptr;
} else if (std::get_if<placeholder_t>(&cachedValue->second)) {
auto attr = root->db->getAttr({cachedValue->first, name});
2020-04-20 14:13:52 +03:00
if (attr) {
if (std::get_if<missing_t>(&attr->second))
return nullptr;
else if (std::get_if<failed_t>(&attr->second)) {
if (forceErrors)
2022-07-12 12:53:24 +03:00
debug("reevaluating failed cached attribute '%s'", getAttrPathStr(name));
else
throw CachedEvalError("cached failure of attribute '%s'", getAttrPathStr(name));
} else
2020-04-20 14:13:52 +03:00
return std::make_shared<AttrCursor>(root,
std::make_pair(shared_from_this(), name), nullptr, std::move(attr));
}
// Incomplete attrset, so need to fall thru and
// evaluate to see whether 'name' exists
} else
return nullptr;
//throw TypeError("'%s' is not an attribute set", getAttrPathStr());
}
}
auto & v = forceValue();
if (v.type() != nAttrs)
2020-04-20 14:13:52 +03:00
return nullptr;
//throw TypeError("'%s' is not an attribute set", getAttrPathStr());
auto attr = v.attrs->get(name);
2019-06-07 23:38:39 +03:00
2020-04-20 14:13:52 +03:00
if (!attr) {
if (root->db) {
if (!cachedValue)
cachedValue = {root->db->setPlaceholder(getKey()), placeholder_t()};
root->db->setMissing({cachedValue->first, name});
}
return nullptr;
}
2020-04-20 14:13:52 +03:00
std::optional<std::pair<AttrId, AttrValue>> cachedValue2;
if (root->db) {
if (!cachedValue)
cachedValue = {root->db->setPlaceholder(getKey()), placeholder_t()};
cachedValue2 = {root->db->setPlaceholder({cachedValue->first, name}), placeholder_t()};
}
return make_ref<AttrCursor>(
2020-04-20 14:13:52 +03:00
root, std::make_pair(shared_from_this(), name), attr->value, std::move(cachedValue2));
}
std::shared_ptr<AttrCursor> AttrCursor::maybeGetAttr(std::string_view name)
{
return maybeGetAttr(root->state.symbols.create(name));
}
ref<AttrCursor> AttrCursor::getAttr(Symbol name, bool forceErrors)
2020-04-20 14:13:52 +03:00
{
auto p = maybeGetAttr(name, forceErrors);
2020-04-20 14:13:52 +03:00
if (!p)
throw Error("attribute '%s' does not exist", getAttrPathStr(name));
return ref(p);
2020-04-20 14:13:52 +03:00
}
ref<AttrCursor> AttrCursor::getAttr(std::string_view name)
2020-04-20 14:13:52 +03:00
{
return getAttr(root->state.symbols.create(name));
}
OrSuggestions<ref<AttrCursor>> AttrCursor::findAlongAttrPath(const std::vector<Symbol> & attrPath, bool force)
2020-04-20 14:13:52 +03:00
{
auto res = shared_from_this();
for (auto & attr : attrPath) {
auto child = res->maybeGetAttr(attr, force);
if (!child) {
auto suggestions = res->getSuggestionsForAttr(attr);
return OrSuggestions<ref<AttrCursor>>::failed(suggestions);
}
res = child;
2020-04-20 14:13:52 +03:00
}
return ref(res);
2020-04-20 14:13:52 +03:00
}
std::string AttrCursor::getString()
{
if (root->db) {
if (!cachedValue)
cachedValue = root->db->getAttr(getKey());
2020-04-20 14:13:52 +03:00
if (cachedValue && !std::get_if<placeholder_t>(&cachedValue->second)) {
2020-06-29 20:08:37 +03:00
if (auto s = std::get_if<string_t>(&cachedValue->second)) {
2020-04-20 14:13:52 +03:00
debug("using cached string attribute '%s'", getAttrPathStr());
2020-06-29 20:08:37 +03:00
return s->first;
2020-04-20 14:13:52 +03:00
} else
root->state.error("'%s' is not a string", getAttrPathStr()).debugThrow<TypeError>();
2020-04-20 14:13:52 +03:00
}
}
auto & v = forceValue();
if (v.type() != nString && v.type() != nPath)
root->state.error("'%s' is not a string but %s", getAttrPathStr()).debugThrow<TypeError>();
2020-04-20 14:13:52 +03:00
return v.type() == nString ? v.c_str() : v.path().to_string();
2020-04-20 14:13:52 +03:00
}
2020-06-29 20:08:37 +03:00
string_t AttrCursor::getStringWithContext()
{
if (root->db) {
if (!cachedValue)
cachedValue = root->db->getAttr(getKey());
2020-06-29 20:08:37 +03:00
if (cachedValue && !std::get_if<placeholder_t>(&cachedValue->second)) {
if (auto s = std::get_if<string_t>(&cachedValue->second)) {
bool valid = true;
for (auto & c : s->second) {
const StorePath & path = std::visit(overloaded {
[&](const NixStringContextElem::DrvDeep & d) -> const StorePath & {
return d.drvPath;
},
[&](const NixStringContextElem::Built & b) -> const StorePath & {
return b.drvPath->getBaseStorePath();
},
[&](const NixStringContextElem::Opaque & o) -> const StorePath & {
return o.path;
},
}, c.raw);
if (!root->state.store->isValidPath(path)) {
valid = false;
break;
}
}
if (valid) {
debug("using cached string attribute '%s'", getAttrPathStr());
return *s;
}
2020-06-29 20:08:37 +03:00
} else
root->state.error("'%s' is not a string", getAttrPathStr()).debugThrow<TypeError>();
2020-06-29 20:08:37 +03:00
}
}
auto & v = forceValue();
Use `std::set<StringContextElem>` not `PathSet` for string contexts Motivation `PathSet` is not correct because string contexts have other forms (`Built` and `DrvDeep`) that are not rendered as plain store paths. Instead of wrongly using `PathSet`, or "stringly typed" using `StringSet`, use `std::std<StringContextElem>`. ----- In support of this change, `NixStringContext` is now defined as `std::std<StringContextElem>` not `std:vector<StringContextElem>`. The old definition was just used by a `getContext` method which was only used by the eval cache. It can be deleted altogether since the types are now unified and the preexisting `copyContext` function already suffices. Summarizing the previous paragraph: Old: - `value/context.hh`: `NixStringContext = std::vector<StringContextElem>` - `value.hh`: `NixStringContext Value::getContext(...)` - `value.hh`: `copyContext(...)` New: - `value/context.hh`: `NixStringContext = std::set<StringContextElem>` - `value.hh`: `copyContext(...)` ---- The string representation of string context elements no longer contains the store dir. The diff of `src/libexpr/tests/value/context.cc` should make clear what the new representation is, so we recommend reviewing that file first. This was done for two reasons: Less API churn: `Value::mkString` and friends did not take a `Store` before. But if `NixStringContextElem::{parse, to_string}` *do* take a store (as they did before), then we cannot have the `Value` functions use them (in order to work with the fully-structured `NixStringContext`) without adding that argument. That would have been a lot of churn of threading the store, and this diff is already large enough, so the easier and less invasive thing to do was simply make the element `parse` and `to_string` functions not take the `Store` reference, and the easiest way to do that was to simply drop the store dir. Space usage: Dropping the `/nix/store/` (or similar) from the internal representation will safe space in the heap of the Nix programming being interpreted. If the heap contains many strings with non-trivial contexts, the saving could add up to something significant. ---- The eval cache version is bumped. The eval cache serialization uses `NixStringContextElem::{parse, to_string}`, and since those functions are changed per the above, that means the on-disk representation is also changed. This is simply done by changing the name of the used for the eval cache from `eval-cache-v4` to eval-cache-v5`. ---- To avoid some duplication `EvalCache::mkPathString` is added to abstract over the simple case of turning a store path to a string with just that string in the context. Context This PR picks up where #7543 left off. That one introduced the fully structured `NixStringContextElem` data type, but kept `PathSet context` as an awkward middle ground between internal `char[][]` interpreter heap string contexts and `NixStringContext` fully parsed string contexts. The infelicity of `PathSet context` was specifically called out during Nix team group review, but it was agreeing that fixing it could be left as future work. This is that future work. A possible follow-up step would be to get rid of the `char[][]` evaluator heap representation, too, but it is not yet clear how to do that. To use `NixStringContextElem` there we would need to get the STL containers to GC pointers in the GC build, and I am not sure how to do that. ---- PR #7543 effectively is writing the inverse of a `mkPathString`, `mkOutputString`, and one more such function for the `DrvDeep` case. I would like that PR to have property tests ensuring it is actually the inverse as expected. This PR sets things up nicely so that reworking that PR to be in that more elegant and better tested way is possible. Co-authored-by: Théophane Hufschmitt <7226587+thufschmitt@users.noreply.github.com>
2023-01-29 03:31:10 +02:00
if (v.type() == nString) {
NixStringContext context;
copyContext(v, context);
return {v.c_str(), std::move(context)};
2023-04-25 17:52:02 +03:00
}
else if (v.type() == nPath)
return {v.path().to_string(), {}};
2020-06-29 20:08:37 +03:00
else
root->state.error("'%s' is not a string but %s", getAttrPathStr()).debugThrow<TypeError>();
2020-06-29 20:08:37 +03:00
}
bool AttrCursor::getBool()
{
if (root->db) {
if (!cachedValue)
cachedValue = root->db->getAttr(getKey());
if (cachedValue && !std::get_if<placeholder_t>(&cachedValue->second)) {
if (auto b = std::get_if<bool>(&cachedValue->second)) {
debug("using cached Boolean attribute '%s'", getAttrPathStr());
return *b;
} else
root->state.error("'%s' is not a Boolean", getAttrPathStr()).debugThrow<TypeError>();
}
}
auto & v = forceValue();
if (v.type() != nBool)
root->state.error("'%s' is not a Boolean", getAttrPathStr()).debugThrow<TypeError>();
return v.boolean;
}
2022-05-13 23:02:28 +03:00
NixInt AttrCursor::getInt()
{
if (root->db) {
if (!cachedValue)
cachedValue = root->db->getAttr(getKey());
if (cachedValue && !std::get_if<placeholder_t>(&cachedValue->second)) {
if (auto i = std::get_if<int_t>(&cachedValue->second)) {
debug("using cached integer attribute '%s'", getAttrPathStr());
return i->x;
2022-05-13 23:02:28 +03:00
} else
throw TypeError("'%s' is not an integer", getAttrPathStr());
2022-05-13 23:02:28 +03:00
}
}
auto & v = forceValue();
if (v.type() != nInt)
throw TypeError("'%s' is not an integer", getAttrPathStr());
2022-05-13 23:02:28 +03:00
return v.integer;
}
std::vector<std::string> AttrCursor::getListOfStrings()
{
if (root->db) {
if (!cachedValue)
cachedValue = root->db->getAttr(getKey());
if (cachedValue && !std::get_if<placeholder_t>(&cachedValue->second)) {
if (auto l = std::get_if<std::vector<std::string>>(&cachedValue->second)) {
debug("using cached list of strings attribute '%s'", getAttrPathStr());
return *l;
} else
throw TypeError("'%s' is not a list of strings", getAttrPathStr());
}
}
debug("evaluating uncached attribute '%s'", getAttrPathStr());
auto & v = getValue();
root->state.forceValue(v, noPos);
if (v.type() != nList)
throw TypeError("'%s' is not a list", getAttrPathStr());
std::vector<std::string> res;
for (auto & elem : v.listItems())
res.push_back(std::string(root->state.forceStringNoCtx(*elem, noPos, "while evaluating an attribute for caching")));
2022-05-04 22:13:49 +03:00
if (root->db)
cachedValue = {root->db->setListOfStrings(getKey(), res), res};
return res;
}
2020-04-20 14:13:52 +03:00
std::vector<Symbol> AttrCursor::getAttrs()
{
if (root->db) {
if (!cachedValue)
cachedValue = root->db->getAttr(getKey());
2020-04-20 14:13:52 +03:00
if (cachedValue && !std::get_if<placeholder_t>(&cachedValue->second)) {
if (auto attrs = std::get_if<std::vector<Symbol>>(&cachedValue->second)) {
debug("using cached attrset attribute '%s'", getAttrPathStr());
return *attrs;
} else
root->state.error("'%s' is not an attribute set", getAttrPathStr()).debugThrow<TypeError>();
2020-04-20 14:13:52 +03:00
}
}
2020-04-20 14:13:52 +03:00
auto & v = forceValue();
if (v.type() != nAttrs)
root->state.error("'%s' is not an attribute set", getAttrPathStr()).debugThrow<TypeError>();
2020-04-20 14:13:52 +03:00
std::vector<Symbol> attrs;
for (auto & attr : *getValue().attrs)
attrs.push_back(attr.name);
std::sort(attrs.begin(), attrs.end(), [&](Symbol a, Symbol b) {
std::string_view sa = root->state.symbols[a], sb = root->state.symbols[b];
return sa < sb;
2020-04-20 14:13:52 +03:00
});
2020-04-20 14:13:52 +03:00
if (root->db)
cachedValue = {root->db->setAttrs(getKey(), attrs), attrs};
2020-04-20 14:13:52 +03:00
return attrs;
}
2020-04-20 14:13:52 +03:00
bool AttrCursor::isDerivation()
{
2020-04-20 14:13:52 +03:00
auto aType = maybeGetAttr("type");
return aType && aType->getString() == "derivation";
}
StorePath AttrCursor::forceDerivation()
{
auto aDrvPath = getAttr(root->state.sDrvPath, true);
auto drvPath = root->state.store->parseStorePath(aDrvPath->getString());
if (!root->state.store->isValidPath(drvPath) && !settings.readOnlyMode) {
/* The eval cache contains 'drvPath', but the actual path has
been garbage-collected. So force it to be regenerated. */
aDrvPath->forceValue();
if (!root->state.store->isValidPath(drvPath))
throw Error("don't know how to recreate store derivation '%s'!",
root->state.store->printStorePath(drvPath));
}
return drvPath;
}
}