csv string input/output\n documentation on text format

csv.documentation

+Tarantool supports CSV file input/output.
+CSV is comma separated values, like this:
+example.txt:
+package,method,return value
+fio,pathjoin,string
+csv,load,table
+none,",comma in field", and ""quote""
+
+Commas in fields must be in quotes
+If there are quotes in a field, it must be double-quotes.
+You can set delimiter and quote character:
+    csv.delimiter = ','
+    csv.quote = '"'
+
+csv.iterate = function(readable[, csv_chunk_size])
+--@brief parse csv string by string
+--@param readable must be string or object with method read(num) returns string
+--@param csv_chunk_size (default 4096). Parser will read by csv_chunk_size symbols
+--@return iter function, iterator state
+
+Example:
+    f = require("fio").open("example.txt", { "O_RDONLY"})
+    for tup in csv.iterate(f) do
+        print(tup[1], tup[2], tup[3])
+    end
+Output:
+    package method  return value
+    fio     pathjoin        string
+    csv     load    table
+    none    ,comma in field and "quote"
+
+csv.load = function(readable, csv_chunk_size)
+--@brief parse csv and make table
+--@return table
+
+
+csv.dump = function(t, writable)
+--@brief dumps tuple or table as csv
+--@param writable must be object with method write(string) like file or socket
+--@return there is no writable it returns csv as string
\ No newline at end of file

src/lua/csv.lua

@@ -75,64 +75,112 @@ local iter = function(csvstate)
     end
 end
 
-local csv = {
-    iterate = function(readable, csv_chunk_size)
-        csv_chunk_size = csv_chunk_size or 4096
-        if type(readable.read) ~= "function" then
-            error("Usage: load(object with read method)")
-        end
+local make_readable = function(s)
+    rd = {}
+    rd.val = s
+    rd.read = function(self, cnt)
+        local res = self.val;
+        self.val = ""
+        return res
+    end
+    return rd
+end
 
-        local it = ffi.new('csv_iterator_t[1]')
-        local csv = ffi.new('csv_t[1]')
-        ffi.C.csv_create(csv)
-        ffi.C.csv_iter_create(it, csv)
-
-        return iter, {readable, csv_chunk_size, csv, it}
-    end,
-    load = function(readable, csv_chunk_size)
-        csv_chunk_size = csv_chunk_size or 4096
-        if type(readable.read) ~= "function" then
-            error("Usage: load(object with read method)")
-        end
+local make_writable = function()
+    wr = {}
+    wr.returnstring = ""
+    wr.write = function(self, s)
+        wr.returnstring = wr.returnstring .. s
+    end
+    return wr
+end
 
-        result = {}
-        for tup in csv.iterate(readable, csv_chunk_size) do
-            table.insert(result, tup)
-        end
+local module = {}
 
-        return result
-    end,
-    dump = function(writable, t)
-        if type(writable.write) ~= "function" or type(t) ~= "table" then
-            error("Usage: dump(writable, table)")
-        end
-        local csv = ffi.new('csv_t[1]')
-        ffi.C.csv_create(csv)
-        local bufsz = 256
-        --local buf = ffi.new('char[?]', bufsz)
-        local buf = csv[0].csv_realloc(ffi.cast(ffi.typeof('void *'), 0), bufsz)
-        if type(t[1]) ~= 'table' then
-            t = {t}
-        end
-        for k, line in pairs(t) do
-            local first = true
-            for k2, field in pairs(line) do
-                strf = tostring(field)
-                if (strf:len() + 1) * 2 > bufsz then
-                    bufsz = (strf:len() + 1) * 2
-                    buf = csv[0].csv_realloc(buf, bufsz)
-                end
-                local len = ffi.C.csv_escape_field(csv, strf, buf)
-                if first then
-                    first = false
-                else
-                    writable:write(',')
-                end
-                writable:write(ffi.string(buf, len))
+module.delimiter = ','
+module.quote = '"'
+
+--@brief parse csv string by string
+--@param readable must be string or object with method read(num) returns string
+--@param csv_chunk_size (default 4096). Parser will read by csv_chunk_size symbols
+--@return iter function, iterator state
+module.iterate = function(readable, csv_chunk_size)
+    csv_chunk_size = csv_chunk_size or 4096
+    if type(readable) == "string" then
+        readable = make_readable(readable)
+    end
+    if type(readable.read) ~= "function" then
+        error("Usage: load(object with method read(num) returns string)")
+    end
+
+    local str = readable:read(csv_chunk_size)
+    if not str then
+        error("Usage: load(object with method read(num) returns string)")
+    end
+    local it = ffi.new('csv_iterator_t[1]')
+    local csv = ffi.new('csv_t[1]')
+    ffi.C.csv_create(csv)
+    csv[0].csv_delim = string.byte(module.delimiter)
+    csv[0].csv_quote = string.byte(module.quote)
+    ffi.C.csv_iter_create(it, csv)
+    ffi.C.csv_feed(it, str)
+
+    return iter, {readable, csv_chunk_size, csv, it}
+end
+
+--@brief parse csv and make table
+--@return table
+module.load = function(readable, csv_chunk_size)
+    csv_chunk_size = csv_chunk_size or 4096
+    result = {}
+    for tup in module.iterate(readable, csv_chunk_size) do
+        table.insert(result, tup)
+    end
+
+    return result
+end
+
+--@brief dumps tuple or table as csv
+--@param writable must be object with method write(string) like file or socket
+--@return there is no writable it returns csv as string
+module.dump = function(t, writable)
+    if type(writable) == "nil" then
+        writable = make_writable()
+    end
+    if type(writable.write) ~= "function" or type(t) ~= "table" then
+        error("Usage: dump(writable, table)")
+    end
+    local csv = ffi.new('csv_t[1]')
+    ffi.C.csv_create(csv)
+    csv[0].csv_delim = string.byte(module.delimiter)
+    csv[0].csv_quote = string.byte(module.quote)
+    local bufsz = 256
+    local buf = csv[0].csv_realloc(ffi.cast(ffi.typeof('void *'), 0), bufsz)
+    if type(t[1]) ~= 'table' then
+        t = {t}
+    end
+    for k, line in pairs(t) do
+        local first = true
+        for k2, field in pairs(line) do
+            strf = tostring(field)
+            if (strf:len() + 1) * 2 > bufsz then
+                bufsz = (strf:len() + 1) * 2
+                buf = csv[0].csv_realloc(buf, bufsz)
+            end
+            local len = ffi.C.csv_escape_field(csv, strf, buf)
+            if first then
+                first = false
+            else
+                writable:write(module.delimiter)
             end
-            writable:write('\n')
+            writable:write(ffi.string(buf, len))
         end
-        csv[0].csv_realloc(buf, 0)
+        writable:write('\n')
     end
-}
-return csv
+    csv[0].csv_realloc(buf, 0)
+    if writable.returnstring then
+        return writable.returnstring
+    end
+end
+
+return module

test/app/csv.result

@@ -51,3 +51,18 @@ aba
 local res = internal.pwrite(self.fh
 iflag = bit.bor(iflag
 
+test str dump:
+quote"" d,",and, comma","both "" of "" t,h,e,m"
+"""""",","","""
+"mul
+ti
+li
+ne
+
+",field
+
+""
+"
+"
+
+test load(dump(t)): true

test/app/csv.test.lua

@@ -84,7 +84,7 @@ t = {
 }
 
 f = require("fio").open(file3, { "O_WRONLY", "O_TRUNC" , "O_CREAT"}, 0x1FF)
-csv.dump(f, t)
+csv.dump(t, f)
 f:close()
 f = fio.open(file3, {'O_RDONLY'}) 
 t2 = csv.load(f, 5)
@@ -100,6 +100,9 @@ for tup in csv.iterate(f) do
 end
 f:close()
 
+print("test str dump:")
+print(csv.dump(t))
+print("test load(dump(t)): " .. tostring(table2str(t) == table2str(csv.load(csv.dump(t)))))
 
 fio.unlink(file1)
 fio.unlink(file2)