module Netmime:sig
..end
Contents
The tutorial has been moved to
Netmime_tut
.typestore =
[ `File of string | `Memory ]
`Memory
means in-memory, `File name
means in the file name
. The body
is stored in decoded form (i.e. without transfer encoding).exception Immutable of string
mime_header_ro
, and mime_body_ro
),
and after that the full class type including write access is defined
(mime_header
, and mime_body
).
The idea is that you can write functions that take an ro value as input to indicate that they do not modify the value. For example:
let number_of_fields (h:#mime_header_ro) =
List.length (h#fields)
This function accepts both mime_header
, and mime_header_ro
values as
input, but the typing ensures that the function cannot mutate anything.
There is another way to ensure that a header or body is not modified.
The read-only flag can be set when creating the object, and this flag
causes that all trials to modify the value will raise the exception
Immutable
. Of course, such trials of mutation are only detected at
run-time.
The advantage of the read-only flag is that it even works if
mutation depends on a condition, but it can be ensured that this
condition is never true. Furthermore, typing is much simpler (getting
subtyping correct can be annoying).
class type mime_header_ro =object
..end
class type mime_header =object
..end
class type mime_body_ro =object
..end
class type mime_body =object
..end
(mime_header, mime_body)
as simple MIME
message with one header and one body. Of course, this simple representation
does not support multi-part messages (attachments). For that reason,
the complex_mime_message
was invented: The body can be further
structured as a sequence of parts that are complex messages themselves.
For example, a mail message with an attachment is usually represented as
(mail_header, `Parts [ (main_header, `Body main_body);
(att_header, `Body att_body) ] )
Here, mail_header
is the real header of the mail message.
main_header
is the header of the main message, usually
only containing the content type of main_body
, the body
of the main message. The attachment has also its own
att_header
, again usually only containing the content type,
and the data of the attachment can be found in att_body
.
Nowadays, mails have often even a more complicated structure
with `Parts
containing nested `Parts
. As complex_mime_message
is recursive, any kind of nesting can be easily represented.
typecomplex_mime_message =
mime_header * complex_mime_body
typecomplex_mime_body =
[ `Body of mime_body | `Parts of complex_mime_message list ]
typecomplex_mime_message_ro =
mime_header_ro * complex_mime_body_ro
typecomplex_mime_body_ro =
[ `Body of mime_body_ro
| `Parts of complex_mime_message_ro list ]
`Parts []
, i.e. `Parts
together with an empty list, is
considered as illegal. Such a value cannot be transformed into
printable text.typemime_message =
mime_header * [ `Body of mime_body ]
typemime_message_ro =
mime_header_ro * [ `Body of mime_body_ro ]
class basic_mime_header :?ro:bool -> (string * string) list ->
mime_header
mime_header
.
class memory_mime_body :?ro:bool -> string ->
mime_body
mime_body
where the value is stored
in-memory.
class file_mime_body :?ro:bool -> ?fin:bool -> string ->
mime_body
mime_body
where the value is stored
in an external file.
val read_mime_header : ?unfold:bool ->
?strip:bool -> ?ro:bool -> Netstream.in_obj_stream -> mime_header
basic_mime_header
.
After returning, the stream is advanced to the byte following the
empty line terminating the header.
Example: To read the header at the beginning of the file "f", use:
let ch = new Netchannels.input_channel (open_in "f") in
let stream = new Netstream.input_stream ch in
let h = read_mime_header stream in
...
stream#close_in(); (* no need to close ch *)
Note that although the stream
position after parsing is exactly
known, the position of ch
cannot be predicted.
unfold
: whether linefeeds are replaced by spaces in the values of the
header fields (Note: defaults to false
here in contrast to
Mimestring.scan_header
!)strip
: whether whitespace at the beginning and at the end of the
header fields is strippedro
: whether the returned header is read-only (default: false)h
into the channel ch
, use
Mimestring.write_header ch h#fields
Link: Mimestring.write_header
typemultipart_style =
[ `Deep | `Flat | `None ]
`None
: Do not handle multipart messages specially. Multipart bodies
are not further decoded, and returned as `Body b
where b
is
the transfer-encoded text representation.`Flat
: If the top-level message is a multipart message, the parts
are separated and returned as list. If the parts are again multipart
messages, these inner multipart messages are not furher decoded
and returned as `Body b
.`Deep
: Multipart messages are recursively decoded and returned as
tree structure.complex_mime_message
structure
is created for a parsed MIME message. `None
means that no parts
are decoded, and messages have always only a simple `Body b
,
even if b
is in reality a multi-part body. With `Flat
, the
top-level multi-part bodies are decoded (if found), and messages
can have a structured `Parts [_, `Body b1; _, `Body b1; ...]
body. Finally, `Deep
allows that inner multi-part bodies are
recursively decoded, and messages can have an arbitrarily complex
form.val decode_mime_body : #mime_header_ro ->
Netchannels.out_obj_channel -> Netchannels.out_obj_channel
let ch' = decode_mime_body hdr ch
:
According to the value of the Content-transfer-encoding header field
in hdr
the encoded MIME body written to ch'
is decoded and transferred
to ch
.
Handles 7bit, 8bit, binary, quoted-printable, base64.
Example: The file "f" contains base64-encoded data, and is to be decoded and to be stored in "g":
let ch_f = new Netchannels.input_channel (open_in "f") in
let ch_g = new Netchannels.output_channel (open_out "g") in
let hdr = new basic_mime_header ["content-transfer-encoding", "base64" ] in
let ch = decode_mime_body hdr ch_g in
ch # output_channel ch_f;
ch # close_out();
ch_g # close_out();
ch_f # close_in();
Note: This function is internally used by read_mime_message
to
decode bodies. There is usually no need to call it directly.
val storage : ?ro:bool ->
?fin:bool -> store -> mime_body * Netchannels.out_obj_channel
store
.
This function can be used to build the storage_style
argument
of the class read_mime_message
(below). For example, this is
useful to store large attachments in external files, as in:
let storage_style hdr =
let filename = hdr ... (* extract from hdr *) in
storage (`File filename)
ro
: whether the returned mime_bodies are read-only or not. Note that
it is always possible to write into the body using the returned
out_obj_channel regardless of the value of ~ro.
Default: falsefin
: whether to finalize bodies stored in files.
Default: falseval read_mime_message : ?unfold:bool ->
?strip:bool ->
?ro:bool ->
?multipart_style:multipart_style ->
?storage_style:(mime_header ->
mime_body * Netchannels.out_obj_channel) ->
Netstream.in_obj_stream -> complex_mime_message
Multipart messages are decoded as specified by multipart_style
(see
above).
Message bodies with content-transfer-encodings of 7bit, 8bit, binary, base64, and quoted-printable can be processed. The bodies are stored without content-transfer-encoding (i.e. in decoded form), but the content-transfer-encoding header field is not removed from the header.
The storage_style
function determines where every message body is
stored. The corresponding header of the body is passed to the function
as argument; the result of the function is a pair of a new mime_body
and an out_obj_channel
writing into this body. You can create such a
pair by calling storage
(above).
By default, the storage_style
is storage ?ro `Memory
for every header.
Here, the designator `Memory
means that the body will be stored in an
O'Caml string. The designator `File fn
would mean that the body will be stored in the
file fn
. The file would be created if it did not yet exist, and
it would be overwritten if it did already exist.
Note that the storage_style
function is called for every non-multipart
body part.
Large message bodies (> maximum string length) are supported if the bodies are stored in files. The memory consumption is optimized for this case, and usually only a small constant amount of memory is needed.
Example:
Parse the MIME message stored in the file f:
let m = read_mime_message
(new input_stream (new input_channel (open_in f)))
unfold
: whether linefeeds are replaced by spaces in the values of the
header fields (Note: defaults to false
here in contrast to
Mimestring.scan_header
!)strip
: whether whitespace at the beginning and at the end of the
header fields is strippedro
: Whether the created MIME headers are read-only or not. Furthermore,
the default storage_style
uses this parameter for the MIME bodies, too.
However, the MIME bodies may have a different read-only flag in general.val encode_mime_body : ?crlf:bool ->
#mime_header_ro ->
Netchannels.out_obj_channel -> Netchannels.out_obj_channel
let ch' = encode_mime_body hdr ch
:
According to the value of the Content-transfer-encoding header field
in hdr
the unencoded MIME body written to ch' is encoded and transferred
to ch.
Handles 7bit, 8bit, binary, quoted-printable, base64.
For an example, see decode_mime_body
which works in a similar way
but performs decoding instead of encoding.
crlf
: if set (this is by default the case) CR/LF will be used for
end-of-line (eol) termination, if not set LF will be used. For 7bit, 8bit and
binary encoding the existing eol delimiters are not rewritten, so this option
has only an effect for quoted-printable and base64.val write_mime_message : ?wr_header:bool ->
?wr_body:bool ->
?nr:int ->
?ret_boundary:string Pervasives.ref ->
?crlf:bool ->
Netchannels.out_obj_channel -> complex_mime_message -> unit
The function fails if multipart messages do not have a multipart content type field (i.e. the content type does not begin with "multipart"). If only the boundary parameter is missing, a good boundary parameter is added to the content type. "Good" means here that it is impossible that the boundary string occurs in the message body if the content-transfer-encoding is quoted-printable or base64, and that such an occurrence is very unlikely if the body is not encoded. If the whole content type field is missing, a "multipart/mixed" type with a boundary parameter is added to the printed header.
Note that already existing boundaries are used, no matter whether they are of good quality or not.
No other header fields are added, deleted or modified. The mentioned modifications are _not_ written back to the passed MIME message but only added to the generated message text.
It is possible in some cases that the boundary does not work (both the existing boundary, and the added boundary). This causes that a wrong and unparseable MIME message is written. In order to ensure a correct MIME message, it is recommended to parse the written text, and to compare the structure of the message trees. It is, however, very unlikely that a problem arises.
Note that if the passed message is a simple message like (_,`Body _), and if no content-transfer-encoding is set, the written message might not end with a linefeed character.
wr_header
: If true, the outermost header is written. Inner headers
of the message parts are written unless ~wr_body=false.wr_body
: If true, the body of the whole message is written; if false,
no body is written at all.nr
: This argument sets the counter that is included in generated
boundaries to a certain minimum value.ret_boundary
: if passed, the boundary of the outermost multipart
message is written to this reference. (Internally used.)crlf
: if set (this is by default the case) CR/LF will be used for
end-of-line (eol) termination, if not set LF will be used. The eol
separator is used for the header, the multipart framing, and for
bodies encoded as quoted-printable or base64. Other eol separators are
left untouched.