Make environment-to-ini support loading key value from file (#24832)

Replace #19857

Close #19856
Close #10311
Close #10123

Major changes:

1. Move a lot of code from `environment-to-ini.go` to `config_env.go` to
make them testable.
2. Add `__FILE` support
3. Update documents
4. Add tests

2 files changed, 239 insertions, 0 deletions

diff --git a/modules/setting/config_env.go b/modules/setting/config_env.go
new file mode 100644
index 0000000000..dca9f2bb47
--- /dev/null
+++ b/modules/setting/config_env.go
@@ -0,0 +1,142 @@
+// Copyright 2023 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package setting
+
+import (
+	"os"
+	"regexp"
+	"strconv"
+	"strings"
+
+	"code.gitea.io/gitea/modules/log"
+
+	"gopkg.in/ini.v1"
+)
+
+const escapeRegexpString = "_0[xX](([0-9a-fA-F][0-9a-fA-F])+)_"
+
+var escapeRegex = regexp.MustCompile(escapeRegexpString)
+
+// decodeEnvSectionKey will decode a portable string encoded Section__Key pair
+// Portable strings are considered to be of the form [A-Z0-9_]*
+// We will encode a disallowed value as the UTF8 byte string preceded by _0X and
+// followed by _. E.g. _0X2C_ for a '-' and _0X2E_ for '.'
+// Section and Key are separated by a plain '__'.
+// The entire section can be encoded as a UTF8 byte string
+func decodeEnvSectionKey(encoded string) (ok bool, section, key string) {
+	inKey := false
+	last := 0
+	escapeStringIndices := escapeRegex.FindAllStringIndex(encoded, -1)
+	for _, unescapeIdx := range escapeStringIndices {
+		preceding := encoded[last:unescapeIdx[0]]
+		if !inKey {
+			if splitter := strings.Index(preceding, "__"); splitter > -1 {
+				section += preceding[:splitter]
+				inKey = true
+				key += preceding[splitter+2:]
+			} else {
+				section += preceding
+			}
+		} else {
+			key += preceding
+		}
+		toDecode := encoded[unescapeIdx[0]+3 : unescapeIdx[1]-1]
+		decodedBytes := make([]byte, len(toDecode)/2)
+		for i := 0; i < len(toDecode)/2; i++ {
+			// Can ignore error here as we know these should be hexadecimal from the regexp
+			byteInt, _ := strconv.ParseInt(toDecode[2*i:2*i+2], 16, 0)
+			decodedBytes[i] = byte(byteInt)
+		}
+		if inKey {
+			key += string(decodedBytes)
+		} else {
+			section += string(decodedBytes)
+		}
+		last = unescapeIdx[1]
+	}
+	remaining := encoded[last:]
+	if !inKey {
+		if splitter := strings.Index(remaining, "__"); splitter > -1 {
+			section += remaining[:splitter]
+			key += remaining[splitter+2:]
+		} else {
+			section += remaining
+		}
+	} else {
+		key += remaining
+	}
+	section = strings.ToLower(section)
+	ok = section != "" && key != ""
+	if !ok {
+		section = ""
+		key = ""
+	}
+	return ok, section, key
+}
+
+// decodeEnvironmentKey decode the environment key to section and key
+// The environment key is in the form of GITEA__SECTION__KEY or GITEA__SECTION__KEY__FILE
+func decodeEnvironmentKey(prefixGitea, suffixFile, envKey string) (ok bool, section, key string, useFileValue bool) {
+	if !strings.HasPrefix(envKey, prefixGitea) {
+		return false, "", "", false
+	}
+	if strings.HasSuffix(envKey, suffixFile) {
+		useFileValue = true
+		envKey = envKey[:len(envKey)-len(suffixFile)]
+	}
+	ok, section, key = decodeEnvSectionKey(envKey[len(prefixGitea):])
+	return ok, section, key, useFileValue
+}
+
+func EnvironmentToConfig(cfg *ini.File, prefixGitea, suffixFile string, envs []string) (changed bool) {
+	for _, kv := range envs {
+		idx := strings.IndexByte(kv, '=')
+		if idx < 0 {
+			continue
+		}
+
+		// parse the environment variable to config section name and key name
+		envKey := kv[:idx]
+		envValue := kv[idx+1:]
+		ok, sectionName, keyName, useFileValue := decodeEnvironmentKey(prefixGitea, suffixFile, envKey)
+		if !ok {
+			continue
+		}
+
+		// use environment value as config value, or read the file content as value if the key indicates a file
+		keyValue := envValue
+		if useFileValue {
+			fileContent, err := os.ReadFile(envValue)
+			if err != nil {
+				log.Error("Error reading file for %s : %v", envKey, envValue, err)
+				continue
+			}
+			keyValue = string(fileContent)
+		}
+
+		// try to set the config value if necessary
+		section, err := cfg.GetSection(sectionName)
+		if err != nil {
+			section, err = cfg.NewSection(sectionName)
+			if err != nil {
+				log.Error("Error creating section: %s : %v", sectionName, err)
+				continue
+			}
+		}
+		key := section.Key(keyName)
+		if key == nil {
+			key, err = section.NewKey(keyName, keyValue)
+			if err != nil {
+				log.Error("Error creating key: %s in section: %s with value: %s : %v", keyName, sectionName, keyValue, err)
+				continue
+			}
+		}
+		oldValue := key.Value()
+		if !changed && oldValue != keyValue {
+			changed = true
+		}
+		key.SetValue(keyValue)
+	}
+	return changed
+}
diff --git a/modules/setting/config_env_test.go b/modules/setting/config_env_test.go
new file mode 100644
index 0000000000..d49464ecf7
--- /dev/null
+++ b/modules/setting/config_env_test.go
@@ -0,0 +1,97 @@
+// Copyright 2023 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package setting
+
+import (
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"gopkg.in/ini.v1"
+)
+
+func TestDecodeEnvSectionKey(t *testing.T) {
+	ok, section, key := decodeEnvSectionKey("SEC__KEY")
+	assert.True(t, ok)
+	assert.Equal(t, "sec", section)
+	assert.Equal(t, "KEY", key)
+
+	ok, section, key = decodeEnvSectionKey("sec__key")
+	assert.True(t, ok)
+	assert.Equal(t, "sec", section)
+	assert.Equal(t, "key", key)
+
+	ok, section, key = decodeEnvSectionKey("LOG_0x2E_CONSOLE__STDERR")
+	assert.True(t, ok)
+	assert.Equal(t, "log.console", section)
+	assert.Equal(t, "STDERR", key)
+
+	ok, section, key = decodeEnvSectionKey("SEC")
+	assert.False(t, ok)
+	assert.Equal(t, "", section)
+	assert.Equal(t, "", key)
+}
+
+func TestDecodeEnvironmentKey(t *testing.T) {
+	prefix := "GITEA__"
+	suffix := "__FILE"
+
+	ok, section, key, file := decodeEnvironmentKey(prefix, suffix, "SEC__KEY")
+	assert.False(t, ok)
+	assert.Equal(t, "", section)
+	assert.Equal(t, "", key)
+	assert.False(t, file)
+
+	ok, section, key, file = decodeEnvironmentKey(prefix, suffix, "GITEA__SEC")
+	assert.False(t, ok)
+	assert.Equal(t, "", section)
+	assert.Equal(t, "", key)
+	assert.False(t, file)
+
+	ok, section, key, file = decodeEnvironmentKey(prefix, suffix, "GITEA__SEC__KEY")
+	assert.True(t, ok)
+	assert.Equal(t, "sec", section)
+	assert.Equal(t, "KEY", key)
+	assert.False(t, file)
+
+	// with "__FILE" suffix, it doesn't support to write "[sec].FILE" to config (no such key FILE is used in Gitea)
+	// but it could be fixed in the future by adding a new suffix like "__VALUE" (no such key VALUE is used in Gitea either)
+	ok, section, key, file = decodeEnvironmentKey(prefix, suffix, "GITEA__SEC__FILE")
+	assert.False(t, ok)
+	assert.Equal(t, "", section)
+	assert.Equal(t, "", key)
+	assert.True(t, file)
+
+	ok, section, key, file = decodeEnvironmentKey(prefix, suffix, "GITEA__SEC__KEY__FILE")
+	assert.True(t, ok)
+	assert.Equal(t, "sec", section)
+	assert.Equal(t, "KEY", key)
+	assert.True(t, file)
+}
+
+func TestEnvironmentToConfig(t *testing.T) {
+	cfg := ini.Empty()
+
+	changed := EnvironmentToConfig(cfg, "GITEA__", "__FILE", nil)
+	assert.False(t, changed)
+
+	cfg, err := ini.Load([]byte(`
+[sec]
+key = old
+`))
+	assert.NoError(t, err)
+
+	changed = EnvironmentToConfig(cfg, "GITEA__", "__FILE", []string{"GITEA__sec__key=new"})
+	assert.True(t, changed)
+	assert.Equal(t, "new", cfg.Section("sec").Key("key").String())
+
+	changed = EnvironmentToConfig(cfg, "GITEA__", "__FILE", []string{"GITEA__sec__key=new"})
+	assert.False(t, changed)
+
+	tmpFile := t.TempDir() + "/the-file"
+	_ = os.WriteFile(tmpFile, []byte("value-from-file"), 0o644)
+	changed = EnvironmentToConfig(cfg, "GITEA__", "__FILE", []string{"GITEA__sec__key__FILE=" + tmpFile})
+	assert.True(t, changed)
+	assert.Equal(t, "value-from-file", cfg.Section("sec").Key("key").String())
+}