gotosocial/vendor/codeberg.org/gruf/go-structr/runtime.go
kim 84279f6a6a
[performance] cache more database calls, reduce required database calls overall (#3290)
* improvements to caching for lists and relationship to accounts / follows

* fix nil panic in AddToList()

* ensure list related caches are correctly invalidated

* ensure returned ID lists are ordered correctly

* bump go-structr to v0.8.9 (returns early if zero uncached keys to be loaded)

* remove zero checks in uncached key load functions (go-structr now handles this)

* fix issues after rebase on upstream/main

* update the expected return order of CSV exports (since list entries are now down by entry creation date)

* rename some funcs, allow deleting list entries for multiple follow IDs at a time, fix up more tests

* use returning statements on delete to get cache invalidation info

* fixes to recent database delete changes

* fix broken list entries delete sql

* remove unused db function

* update remainder of delete functions to behave in similar way, some other small tweaks

* fix delete user sql, allow returning on err no entries

* uncomment + fix list database tests

* update remaining list tests

* update envparsing test

* add comments to each specific key being invalidated

* add more cache invalidation explanatory comments

* whoops; actually delete poll votes from database in the DeletePollByID() func

* remove added but-commented-out field

* improved comment regarding paging being disabled

* make cache invalidation comments match what's actually happening

* fix up delete query comments to match what is happening

* rename function to read a bit better

* don't use ErrNoEntries on delete when not needed (it's only needed for a RETURNING call)

* update function name in test

* move list exclusivity check to AFTER eligibility check. use log.Panic() instead of panic()

* use the poll_id column in poll_votes for selecting votes in poll ID

* fix function name
2024-09-16 16:46:09 +00:00

189 lines
4 KiB
Go

package structr
import (
"fmt"
"reflect"
"unicode"
"unicode/utf8"
"unsafe"
"codeberg.org/gruf/go-mangler"
)
// struct_field contains pre-prepared type
// information about a struct's field member,
// including memory offset and hash function.
type struct_field struct {
rtype reflect.Type
// struct field type mangling
// (i.e. fast serializing) fn.
mangle mangler.Mangler
// zero value data, used when
// nil encountered during ptr
// offset following.
zero unsafe.Pointer
// mangled zero value string,
// if set this indicates zero
// values of field not allowed
zerostr string
// offsets defines whereabouts in
// memory this field is located.
offsets []next_offset
}
// next_offset defines a next offset location
// in a struct_field, first by the number of
// derefences required, then by offset from
// that final memory location.
type next_offset struct {
derefs uint
offset uintptr
}
// find_field will search for a struct field with given set of names,
// where names is a len > 0 slice of names account for struct nesting.
func find_field(t reflect.Type, names []string) (sfield struct_field) {
var (
// is_exported returns whether name is exported
// from a package; can be func or struct field.
is_exported = func(name string) bool {
r, _ := utf8.DecodeRuneInString(name)
return unicode.IsUpper(r)
}
// pop_name pops the next name from
// the provided slice of field names.
pop_name = func() string {
name := names[0]
names = names[1:]
if !is_exported(name) {
panicf("field is not exported: %s", name)
}
return name
}
// field is the iteratively searched
// struct field value in below loop.
field reflect.StructField
)
for len(names) > 0 {
// Pop next name.
name := pop_name()
var off next_offset
// Dereference any ptrs to struct.
for t.Kind() == reflect.Pointer {
t = t.Elem()
off.derefs++
}
// Check for valid struct type.
if t.Kind() != reflect.Struct {
panicf("field %s is not struct (or ptr-to): %s", t, name)
}
var ok bool
// Look for next field by name.
field, ok = t.FieldByName(name)
if !ok {
panicf("unknown field: %s", name)
}
// Set next offset value.
off.offset = field.Offset
sfield.offsets = append(sfield.offsets, off)
// Set the next type.
t = field.Type
}
// Set final type.
sfield.rtype = t
// Find mangler for field type.
sfield.mangle = mangler.Get(t)
// Get new zero value data ptr.
v := reflect.New(t).Elem()
zptr := eface_data(v.Interface())
zstr := sfield.mangle(nil, zptr)
sfield.zerostr = string(zstr)
sfield.zero = zptr
return
}
// extract_fields extracts given structfields from the provided value type,
// this is done using predetermined struct field memory offset locations.
func extract_fields(ptr unsafe.Pointer, fields []struct_field) []unsafe.Pointer {
// Prepare slice of field value pointers.
ptrs := make([]unsafe.Pointer, len(fields))
for i, field := range fields {
// loop scope.
fptr := ptr
for _, offset := range field.offsets {
// Dereference any ptrs to offset.
fptr = deref(fptr, offset.derefs)
if fptr == nil {
break
}
// Jump forward by offset to next ptr.
fptr = unsafe.Pointer(uintptr(fptr) +
offset.offset)
}
if like_ptr(field.rtype) && fptr != nil {
// Further dereference value ptr.
fptr = *(*unsafe.Pointer)(fptr)
}
if fptr == nil {
// Use zero value.
fptr = field.zero
}
// Set field ptr.
ptrs[i] = fptr
}
return ptrs
}
// like_ptr returns whether type's kind is ptr-like.
func like_ptr(t reflect.Type) bool {
switch t.Kind() {
case reflect.Pointer,
reflect.Map,
reflect.Chan,
reflect.Func:
return true
}
return false
}
// deref will dereference ptr 'n' times (or until nil).
func deref(p unsafe.Pointer, n uint) unsafe.Pointer {
for ; n > 0; n-- {
if p == nil {
return nil
}
p = *(*unsafe.Pointer)(p)
}
return p
}
// panicf provides a panic with string formatting.
func panicf(format string, args ...any) {
panic(fmt.Sprintf(format, args...))
}