// Package strmangle is a collection of string manipulation functions. // Primarily used by boil and templates for code generation. // Because it is focused on pipelining inside templates // you will see some odd parameter ordering. package strmangle import ( "fmt" "math" "regexp" "sort" "strings" "sync" ) var ( idAlphabet = []byte("abcdefghijklmnopqrstuvwxyz") smartQuoteRgx = regexp.MustCompile(`^(?i)"?[a-z_][_a-z0-9]*"?(\."?[_a-z][_a-z0-9]*"?)*(\.\*)?$`) rgxEnum = regexp.MustCompile(`^enum(\.[a-z_]+)?\((,?'[^']+')+\)$`) rgxEnumIsOK = regexp.MustCompile(`^(?i)[a-z][a-z0-9_]*$`) rgxEnumShouldTitle = regexp.MustCompile(`^[a-z][a-z0-9_]*$`) ) var uppercaseWords = map[string]struct{}{ "acl": {}, "api": {}, "ascii": {}, "cpu": {}, "eof": {}, "guid": {}, "id": {}, "ip": {}, "json": {}, "ram": {}, "sla": {}, "udp": {}, "ui": {}, "uid": {}, "uuid": {}, "uri": {}, "url": {}, "utf8": {}, } var reservedWords = map[string]struct{}{ "break": {}, "case": {}, "chan": {}, "const": {}, "continue": {}, "default": {}, "defer": {}, "else": {}, "fallthrough": {}, "for": {}, "func": {}, "go": {}, "goto": {}, "if": {}, "import": {}, "interface": {}, "map": {}, "package": {}, "range": {}, "return": {}, "select": {}, "struct": {}, "switch": {}, "type": {}, "var": {}, } func init() { // Our Boil inflection Ruleset does not include uncountable inflections. // This way, people using words like Sheep will not have // collisions with their model name (Sheep) and their // function name (Sheep()). Instead, it will // use the regular inflection rules: Sheep, Sheeps(). boilRuleset = newBoilRuleset() } // SchemaTable returns a table name with a schema prefixed if // using a database that supports real schemas, for example, // for Postgres: "schema_name"."table_name", // for MS SQL: [schema_name].[table_name], versus // simply "table_name" for MySQL (because it does not support real schemas) func SchemaTable(lq, rq string, driver string, schema string, table string) string { if (driver == "postgres" && schema != "public") || driver == "mssql" { return fmt.Sprintf(`%s%s%s.%s%s%s`, lq, schema, rq, lq, table, rq) } return fmt.Sprintf(`%s%s%s`, lq, table, rq) } // IdentQuote attempts to quote simple identifiers in SQL tatements func IdentQuote(lq byte, rq byte, s string) string { if strings.ToLower(s) == "null" || s == "?" { return s } if m := smartQuoteRgx.MatchString(s); m != true { return s } buf := GetBuffer() defer PutBuffer(buf) splits := strings.Split(s, ".") for i, split := range splits { if i != 0 { buf.WriteByte('.') } if split[0] == lq || split[len(split)-1] == rq || split == "*" { buf.WriteString(split) continue } buf.WriteByte(lq) buf.WriteString(split) buf.WriteByte(rq) } return buf.String() } // IdentQuoteSlice applies IdentQuote to a slice. func IdentQuoteSlice(lq byte, rq byte, s []string) []string { if len(s) == 0 { return s } strs := make([]string, len(s)) for i, str := range s { strs[i] = IdentQuote(lq, rq, str) } return strs } // Identifier is a base conversion from Base 10 integers to Base 26 // integers that are represented by an alphabet from a-z // See tests for example outputs. func Identifier(in int) string { ln := len(idAlphabet) var n int if in == 0 { n = 1 } else { n = 1 + int(math.Log(float64(in))/math.Log(float64(ln))) } cols := GetBuffer() defer PutBuffer(cols) for i := 0; i < n; i++ { divisor := int(math.Pow(float64(ln), float64(n-i-1))) rem := in / divisor cols.WriteByte(idAlphabet[rem]) in -= rem * divisor } return cols.String() } // QuoteCharacter returns a string that allows the quote character // to be embedded into a Go string that uses double quotes: func QuoteCharacter(q byte) string { if q == '"' { return `\"` } return string(q) } // Plural converts singular words to plural words (eg: person to people) func Plural(name string) string { buf := GetBuffer() defer PutBuffer(buf) splits := strings.Split(name, "_") for i := 0; i < len(splits); i++ { if i != 0 { buf.WriteByte('_') } if i == len(splits)-1 { buf.WriteString(boilRuleset.Pluralize(splits[len(splits)-1])) break } buf.WriteString(splits[i]) } return buf.String() } // Singular converts plural words to singular words (eg: people to person) func Singular(name string) string { buf := GetBuffer() defer PutBuffer(buf) splits := strings.Split(name, "_") for i := 0; i < len(splits); i++ { if i != 0 { buf.WriteByte('_') } if i == len(splits)-1 { buf.WriteString(boilRuleset.Singularize(splits[len(splits)-1])) break } buf.WriteString(splits[i]) } return buf.String() } // titleCaseCache holds the mapping of title cases. // Example: map["MyWord"] == "my_word" var ( mut sync.RWMutex titleCaseCache = map[string]string{} ) // TitleCase changes a snake-case variable name // into a go styled object variable name of "ColumnName". // titleCase also fully uppercases "ID" components of names, for example // "column_name_id" to "ColumnNameID". // // Note: This method is ugly because it has been highly optimized, // we found that it was a fairly large bottleneck when we were using regexp. func TitleCase(n string) string { // Attempt to fetch from cache mut.RLock() val, ok := titleCaseCache[n] mut.RUnlock() if ok { return val } ln := len(n) name := []byte(n) buf := GetBuffer() start := 0 end := 0 for start < ln { // Find the start and end of the underscores to account // for the possibility of being multiple underscores in a row. if end < ln { if name[start] == '_' { start++ end++ continue // Once we have found the end of the underscores, we can // find the end of the first full word. } else if name[end] != '_' { end++ continue } } word := name[start:end] wordLen := len(word) var vowels bool numStart := wordLen for i, c := range word { vowels = vowels || (c == 97 || c == 101 || c == 105 || c == 111 || c == 117 || c == 121) if c > 47 && c < 58 && numStart == wordLen { numStart = i } } _, match := uppercaseWords[string(word[:numStart])] if match || !vowels { // Uppercase all a-z characters for _, c := range word { if c > 96 && c < 123 { buf.WriteByte(c - 32) } else { buf.WriteByte(c) } } } else { if c := word[0]; c > 96 && c < 123 { buf.WriteByte(word[0] - 32) buf.Write(word[1:]) } else { buf.Write(word) } } start = end + 1 end = start } ret := buf.String() PutBuffer(buf) // Cache the title case result mut.Lock() titleCaseCache[n] = ret mut.Unlock() return ret } // CamelCase takes a variable name in the format of "var_name" and converts // it into a go styled variable name of "varName". // camelCase also fully uppercases "ID" components of names, for example // "var_name_id" to "varNameID". func CamelCase(name string) string { buf := GetBuffer() defer PutBuffer(buf) index := -1 for i := 0; i < len(name); i++ { if name[i] != '_' { index = i break } } if index != -1 { name = name[index:] } else { return "" } index = -1 for i := 0; i < len(name); i++ { if name[i] == '_' { index = i break } } if index == -1 { buf.WriteString(name) } else { buf.WriteString(name[:index]) buf.WriteString(TitleCase(name[index+1:])) } return buf.String() } // TitleCaseIdentifier splits on dots and then titlecases each fragment. // map titleCase (split c ".") func TitleCaseIdentifier(id string) string { nextDot := strings.IndexByte(id, '.') if nextDot < 0 { return TitleCase(id) } buf := GetBuffer() defer PutBuffer(buf) lastDot := 0 ln := len(id) addDots := false for i := 0; nextDot >= 0; i++ { fragment := id[lastDot:nextDot] titled := TitleCase(fragment) if addDots { buf.WriteByte('.') } buf.WriteString(titled) addDots = true if nextDot == ln { break } lastDot = nextDot + 1 if nextDot = strings.IndexByte(id[lastDot:], '.'); nextDot >= 0 { nextDot += lastDot } else { nextDot = ln } } return buf.String() } // MakeStringMap converts a map[string]string into the format: // "key": "value", "key": "value" func MakeStringMap(types map[string]string) string { buf := GetBuffer() defer PutBuffer(buf) keys := make([]string, 0, len(types)) for k := range types { keys = append(keys, k) } sort.Strings(keys) c := 0 for _, k := range keys { v := types[k] buf.WriteString(fmt.Sprintf("`%s`: `%s`", k, v)) if c < len(types)-1 { buf.WriteString(", ") } c++ } return buf.String() } // StringMap maps a function over a slice of strings. func StringMap(modifier func(string) string, strs []string) []string { ret := make([]string, len(strs)) for i, str := range strs { ret[i] = modifier(str) } return ret } // PrefixStringSlice with the given str. func PrefixStringSlice(str string, strs []string) []string { ret := make([]string, len(strs)) for i, s := range strs { ret[i] = fmt.Sprintf("%s%s", str, s) } return ret } // Placeholders generates the SQL statement placeholders for in queries. // For example, ($1,$2,$3),($4,$5,$6) etc. // It will start counting placeholders at "start". // If indexPlaceholders is false, it will convert to ? instead of $1 etc. func Placeholders(indexPlaceholders bool, count int, start int, group int) string { buf := GetBuffer() defer PutBuffer(buf) if start == 0 || group == 0 { panic("Invalid start or group numbers supplied.") } if group > 1 { buf.WriteByte('(') } for i := 0; i < count; i++ { if i != 0 { if group > 1 && i%group == 0 { buf.WriteString("),(") } else { buf.WriteByte(',') } } if indexPlaceholders { buf.WriteString(fmt.Sprintf("$%d", start+i)) } else { buf.WriteByte('?') } } if group > 1 { buf.WriteByte(')') } return buf.String() } // SetParamNames takes a slice of columns and returns a comma separated // list of parameter names for a template statement SET clause. // eg: "col1"=$1, "col2"=$2, "col3"=$3 func SetParamNames(lq, rq string, start int, columns []string) string { buf := GetBuffer() defer PutBuffer(buf) for i, c := range columns { if start != 0 { buf.WriteString(fmt.Sprintf(`%s%s%s=$%d`, lq, c, rq, i+start)) } else { buf.WriteString(fmt.Sprintf(`%s%s%s=?`, lq, c, rq)) } if i < len(columns)-1 { buf.WriteByte(',') } } return buf.String() } // WhereClause returns the where clause using start as the $ flag index // For example, if start was 2 output would be: "colthing=$2 AND colstuff=$3" func WhereClause(lq, rq string, start int, cols []string) string { buf := GetBuffer() defer PutBuffer(buf) for i, c := range cols { if start != 0 { buf.WriteString(fmt.Sprintf(`%s%s%s=$%d`, lq, c, rq, start+i)) } else { buf.WriteString(fmt.Sprintf(`%s%s%s=?`, lq, c, rq)) } if i < len(cols)-1 { buf.WriteString(" AND ") } } return buf.String() } // JoinSlices merges two string slices of equal length func JoinSlices(sep string, a, b []string) []string { lna, lnb := len(a), len(b) if lna != lnb { panic("joinSlices: can only merge slices of same length") } else if lna == 0 { return nil } ret := make([]string, len(a)) for i, elem := range a { ret[i] = fmt.Sprintf("%s%s%s", elem, sep, b[i]) } return ret } // StringSliceMatch returns true if the length of both // slices is the same, and the elements of both slices are the same. // The elements can be in any order. func StringSliceMatch(a []string, b []string) bool { if len(a) != len(b) { return false } for _, aval := range a { found := false for _, bval := range b { if bval == aval { found = true break } } if !found { return false } } return true } // ContainsAny returns true if any of the passed in strings are // found in the passed in string slice func ContainsAny(a []string, finds ...string) bool { for _, s := range a { for _, find := range finds { if s == find { return true } } } return false } // GenerateTags converts a slice of tag strings into tags that // can be passed onto the end of a struct, for example: // tags: ["xml", "db"] convert to: xml:"column_name" db:"column_name" func GenerateTags(tags []string, columnName string) string { buf := GetBuffer() defer PutBuffer(buf) for _, tag := range tags { buf.WriteString(tag) buf.WriteString(`:"`) buf.WriteString(columnName) buf.WriteString(`" `) } return buf.String() } // GenerateIgnoreTags converts a slice of tag strings into // ignore tags that can be passed onto the end of a struct, for example: // tags: ["xml", "db"] convert to: xml:"-" db:"-" func GenerateIgnoreTags(tags []string) string { buf := GetBuffer() defer PutBuffer(buf) for _, tag := range tags { buf.WriteString(tag) buf.WriteString(`:"-" `) } return buf.String() } // ParseEnumVals returns the values from an enum string // // Postgres and MySQL drivers return different values // psql: enum.enum_name('values'...) // mysql: enum('values'...) func ParseEnumVals(s string) []string { if !rgxEnum.MatchString(s) { return nil } startIndex := strings.IndexByte(s, '(') s = s[startIndex+2 : len(s)-2] return strings.Split(s, "','") } // ParseEnumName returns the name portion of an enum if it exists // // Postgres and MySQL drivers return different values // psql: enum.enum_name('values'...) // mysql: enum('values'...) // In the case of mysql, the name will never return anything func ParseEnumName(s string) string { if !rgxEnum.MatchString(s) { return "" } endIndex := strings.IndexByte(s, '(') s = s[:endIndex] startIndex := strings.IndexByte(s, '.') if startIndex < 0 { return "" } return s[startIndex+1:] } // IsEnumNormal checks a set of eval values to see if they're "normal" func IsEnumNormal(values []string) bool { for _, v := range values { if !rgxEnumIsOK.MatchString(v) { return false } } return true } // ShouldTitleCaseEnum checks a value to see if it's title-case-able func ShouldTitleCaseEnum(value string) bool { return rgxEnumShouldTitle.MatchString(value) } // ReplaceReservedWords takes a word and replaces it with word_ if it's found // in the list of reserved words. func ReplaceReservedWords(word string) string { if _, ok := reservedWords[word]; ok { return word + "_" } return word }