mirror of
https://git.notmuchmail.org/git/notmuch
synced 2024-12-23 01:44:52 +01:00
0e63372efe
This option allows the caller to suppress the output of the bodies of the messages. Currently this is only implemented for format=json. This is used by notmuch-pick.el (although not needed) because it gives a speed-up of at least a factor of a two (and in some cases a speed up of more than a factor of 8); moreover it reduces the memory usage in emacs hugely.
1162 lines
33 KiB
C
1162 lines
33 KiB
C
/* notmuch - Not much of an email program, (just index and search)
|
|
*
|
|
* Copyright © 2009 Carl Worth
|
|
*
|
|
* This program is free software: you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation, either version 3 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program. If not, see http://www.gnu.org/licenses/ .
|
|
*
|
|
* Author: Carl Worth <cworth@cworth.org>
|
|
*/
|
|
|
|
#include "notmuch-client.h"
|
|
#include "gmime-filter-reply.h"
|
|
|
|
static notmuch_status_t
|
|
format_part_text (const void *ctx, mime_node_t *node,
|
|
int indent, const notmuch_show_params_t *params);
|
|
|
|
static const notmuch_show_format_t format_text = {
|
|
.part = format_part_text,
|
|
};
|
|
|
|
static notmuch_status_t
|
|
format_part_json_entry (const void *ctx, mime_node_t *node,
|
|
int indent, const notmuch_show_params_t *params);
|
|
|
|
static const notmuch_show_format_t format_json = {
|
|
.message_set_start = "[",
|
|
.part = format_part_json_entry,
|
|
.message_set_sep = ", ",
|
|
.message_set_end = "]",
|
|
.null_message = "null"
|
|
};
|
|
|
|
static notmuch_status_t
|
|
format_part_mbox (const void *ctx, mime_node_t *node,
|
|
int indent, const notmuch_show_params_t *params);
|
|
|
|
static const notmuch_show_format_t format_mbox = {
|
|
.part = format_part_mbox,
|
|
};
|
|
|
|
static notmuch_status_t
|
|
format_part_raw (unused (const void *ctx), mime_node_t *node,
|
|
unused (int indent),
|
|
unused (const notmuch_show_params_t *params));
|
|
|
|
static const notmuch_show_format_t format_raw = {
|
|
.part = format_part_raw,
|
|
};
|
|
|
|
static const char *
|
|
_get_tags_as_string (const void *ctx, notmuch_message_t *message)
|
|
{
|
|
notmuch_tags_t *tags;
|
|
int first = 1;
|
|
const char *tag;
|
|
char *result;
|
|
|
|
result = talloc_strdup (ctx, "");
|
|
if (result == NULL)
|
|
return NULL;
|
|
|
|
for (tags = notmuch_message_get_tags (message);
|
|
notmuch_tags_valid (tags);
|
|
notmuch_tags_move_to_next (tags))
|
|
{
|
|
tag = notmuch_tags_get (tags);
|
|
|
|
result = talloc_asprintf_append (result, "%s%s",
|
|
first ? "" : " ", tag);
|
|
first = 0;
|
|
}
|
|
|
|
return result;
|
|
}
|
|
|
|
/* Get a nice, single-line summary of message. */
|
|
static const char *
|
|
_get_one_line_summary (const void *ctx, notmuch_message_t *message)
|
|
{
|
|
const char *from;
|
|
time_t date;
|
|
const char *relative_date;
|
|
const char *tags;
|
|
|
|
from = notmuch_message_get_header (message, "from");
|
|
|
|
date = notmuch_message_get_date (message);
|
|
relative_date = notmuch_time_relative_date (ctx, date);
|
|
|
|
tags = _get_tags_as_string (ctx, message);
|
|
|
|
return talloc_asprintf (ctx, "%s (%s) (%s)",
|
|
from, relative_date, tags);
|
|
}
|
|
|
|
static void
|
|
format_message_json (const void *ctx, notmuch_message_t *message)
|
|
{
|
|
notmuch_tags_t *tags;
|
|
int first = 1;
|
|
void *ctx_quote = talloc_new (ctx);
|
|
time_t date;
|
|
const char *relative_date;
|
|
|
|
date = notmuch_message_get_date (message);
|
|
relative_date = notmuch_time_relative_date (ctx, date);
|
|
|
|
printf ("\"id\": %s, \"match\": %s, \"excluded\": %s, \"filename\": %s, \"timestamp\": %ld, \"date_relative\": \"%s\", \"tags\": [",
|
|
json_quote_str (ctx_quote, notmuch_message_get_message_id (message)),
|
|
notmuch_message_get_flag (message, NOTMUCH_MESSAGE_FLAG_MATCH) ? "true" : "false",
|
|
notmuch_message_get_flag (message, NOTMUCH_MESSAGE_FLAG_EXCLUDED) ? "true" : "false",
|
|
json_quote_str (ctx_quote, notmuch_message_get_filename (message)),
|
|
date, relative_date);
|
|
|
|
for (tags = notmuch_message_get_tags (message);
|
|
notmuch_tags_valid (tags);
|
|
notmuch_tags_move_to_next (tags))
|
|
{
|
|
printf("%s%s", first ? "" : ",",
|
|
json_quote_str (ctx_quote, notmuch_tags_get (tags)));
|
|
first = 0;
|
|
}
|
|
printf("], ");
|
|
talloc_free (ctx_quote);
|
|
}
|
|
|
|
/* Extract just the email address from the contents of a From:
|
|
* header. */
|
|
static const char *
|
|
_extract_email_address (const void *ctx, const char *from)
|
|
{
|
|
InternetAddressList *addresses;
|
|
InternetAddress *address;
|
|
InternetAddressMailbox *mailbox;
|
|
const char *email = "MAILER-DAEMON";
|
|
|
|
addresses = internet_address_list_parse_string (from);
|
|
|
|
/* Bail if there is no address here. */
|
|
if (addresses == NULL || internet_address_list_length (addresses) < 1)
|
|
goto DONE;
|
|
|
|
/* Otherwise, just use the first address. */
|
|
address = internet_address_list_get_address (addresses, 0);
|
|
|
|
/* The From header should never contain an address group rather
|
|
* than a mailbox. So bail if it does. */
|
|
if (! INTERNET_ADDRESS_IS_MAILBOX (address))
|
|
goto DONE;
|
|
|
|
mailbox = INTERNET_ADDRESS_MAILBOX (address);
|
|
email = internet_address_mailbox_get_addr (mailbox);
|
|
email = talloc_strdup (ctx, email);
|
|
|
|
DONE:
|
|
if (addresses)
|
|
g_object_unref (addresses);
|
|
|
|
return email;
|
|
}
|
|
|
|
/* Return 1 if 'line' is an mbox From_ line---that is, a line
|
|
* beginning with zero or more '>' characters followed by the
|
|
* characters 'F', 'r', 'o', 'm', and space.
|
|
*
|
|
* Any characters at all may appear after that in the line.
|
|
*/
|
|
static int
|
|
_is_from_line (const char *line)
|
|
{
|
|
const char *s = line;
|
|
|
|
if (line == NULL)
|
|
return 0;
|
|
|
|
while (*s == '>')
|
|
s++;
|
|
|
|
if (STRNCMP_LITERAL (s, "From ") == 0)
|
|
return 1;
|
|
else
|
|
return 0;
|
|
}
|
|
|
|
void
|
|
format_headers_json (const void *ctx, GMimeMessage *message, notmuch_bool_t reply)
|
|
{
|
|
void *local = talloc_new (ctx);
|
|
InternetAddressList *recipients;
|
|
const char *recipients_string;
|
|
|
|
printf ("{%s: %s",
|
|
json_quote_str (local, "Subject"),
|
|
json_quote_str (local, g_mime_message_get_subject (message)));
|
|
printf (", %s: %s",
|
|
json_quote_str (local, "From"),
|
|
json_quote_str (local, g_mime_message_get_sender (message)));
|
|
recipients = g_mime_message_get_recipients (message, GMIME_RECIPIENT_TYPE_TO);
|
|
recipients_string = internet_address_list_to_string (recipients, 0);
|
|
if (recipients_string)
|
|
printf (", %s: %s",
|
|
json_quote_str (local, "To"),
|
|
json_quote_str (local, recipients_string));
|
|
recipients = g_mime_message_get_recipients (message, GMIME_RECIPIENT_TYPE_CC);
|
|
recipients_string = internet_address_list_to_string (recipients, 0);
|
|
if (recipients_string)
|
|
printf (", %s: %s",
|
|
json_quote_str (local, "Cc"),
|
|
json_quote_str (local, recipients_string));
|
|
|
|
if (reply) {
|
|
printf (", %s: %s",
|
|
json_quote_str (local, "In-reply-to"),
|
|
json_quote_str (local, g_mime_object_get_header (GMIME_OBJECT (message), "In-reply-to")));
|
|
|
|
printf (", %s: %s",
|
|
json_quote_str (local, "References"),
|
|
json_quote_str (local, g_mime_object_get_header (GMIME_OBJECT (message), "References")));
|
|
} else {
|
|
printf (", %s: %s",
|
|
json_quote_str (local, "Date"),
|
|
json_quote_str (local, g_mime_message_get_date_as_string (message)));
|
|
}
|
|
|
|
printf ("}");
|
|
|
|
talloc_free (local);
|
|
}
|
|
|
|
/* Write a MIME text part out to the given stream.
|
|
*
|
|
* If (flags & NOTMUCH_SHOW_TEXT_PART_REPLY), this prepends "> " to
|
|
* each output line.
|
|
*
|
|
* Both line-ending conversion (CRLF->LF) and charset conversion ( ->
|
|
* UTF-8) will be performed, so it is inappropriate to call this
|
|
* function with a non-text part. Doing so will trigger an internal
|
|
* error.
|
|
*/
|
|
void
|
|
show_text_part_content (GMimeObject *part, GMimeStream *stream_out,
|
|
notmuch_show_text_part_flags flags)
|
|
{
|
|
GMimeContentType *content_type = g_mime_object_get_content_type (GMIME_OBJECT (part));
|
|
GMimeStream *stream_filter = NULL;
|
|
GMimeDataWrapper *wrapper;
|
|
const char *charset;
|
|
|
|
if (! g_mime_content_type_is_type (content_type, "text", "*"))
|
|
INTERNAL_ERROR ("Illegal request to format non-text part (%s) as text.",
|
|
g_mime_content_type_to_string (content_type));
|
|
|
|
if (stream_out == NULL)
|
|
return;
|
|
|
|
stream_filter = g_mime_stream_filter_new (stream_out);
|
|
g_mime_stream_filter_add(GMIME_STREAM_FILTER (stream_filter),
|
|
g_mime_filter_crlf_new (FALSE, FALSE));
|
|
|
|
charset = g_mime_object_get_content_type_parameter (part, "charset");
|
|
if (charset) {
|
|
GMimeFilter *charset_filter;
|
|
charset_filter = g_mime_filter_charset_new (charset, "UTF-8");
|
|
/* This result can be NULL for things like "unknown-8bit".
|
|
* Don't set a NULL filter as that makes GMime print
|
|
* annoying assertion-failure messages on stderr. */
|
|
if (charset_filter) {
|
|
g_mime_stream_filter_add (GMIME_STREAM_FILTER (stream_filter),
|
|
charset_filter);
|
|
g_object_unref (charset_filter);
|
|
}
|
|
|
|
}
|
|
|
|
if (flags & NOTMUCH_SHOW_TEXT_PART_REPLY) {
|
|
GMimeFilter *reply_filter;
|
|
reply_filter = g_mime_filter_reply_new (TRUE);
|
|
if (reply_filter) {
|
|
g_mime_stream_filter_add (GMIME_STREAM_FILTER (stream_filter),
|
|
reply_filter);
|
|
g_object_unref (reply_filter);
|
|
}
|
|
}
|
|
|
|
wrapper = g_mime_part_get_content_object (GMIME_PART (part));
|
|
if (wrapper && stream_filter)
|
|
g_mime_data_wrapper_write_to_stream (wrapper, stream_filter);
|
|
if (stream_filter)
|
|
g_object_unref(stream_filter);
|
|
}
|
|
|
|
#ifdef GMIME_ATLEAST_26
|
|
static const char*
|
|
signature_status_to_string (GMimeSignatureStatus x)
|
|
{
|
|
switch (x) {
|
|
case GMIME_SIGNATURE_STATUS_GOOD:
|
|
return "good";
|
|
case GMIME_SIGNATURE_STATUS_BAD:
|
|
return "bad";
|
|
case GMIME_SIGNATURE_STATUS_ERROR:
|
|
return "error";
|
|
}
|
|
return "unknown";
|
|
}
|
|
#else
|
|
static const char*
|
|
signer_status_to_string (GMimeSignerStatus x)
|
|
{
|
|
switch (x) {
|
|
case GMIME_SIGNER_STATUS_NONE:
|
|
return "none";
|
|
case GMIME_SIGNER_STATUS_GOOD:
|
|
return "good";
|
|
case GMIME_SIGNER_STATUS_BAD:
|
|
return "bad";
|
|
case GMIME_SIGNER_STATUS_ERROR:
|
|
return "error";
|
|
}
|
|
return "unknown";
|
|
}
|
|
#endif
|
|
|
|
#ifdef GMIME_ATLEAST_26
|
|
static void
|
|
format_part_sigstatus_json (mime_node_t *node)
|
|
{
|
|
GMimeSignatureList *siglist = node->sig_list;
|
|
|
|
printf ("[");
|
|
|
|
if (!siglist) {
|
|
printf ("]");
|
|
return;
|
|
}
|
|
|
|
void *ctx_quote = talloc_new (NULL);
|
|
int i;
|
|
for (i = 0; i < g_mime_signature_list_length (siglist); i++) {
|
|
GMimeSignature *signature = g_mime_signature_list_get_signature (siglist, i);
|
|
|
|
if (i > 0)
|
|
printf (", ");
|
|
|
|
printf ("{");
|
|
|
|
/* status */
|
|
GMimeSignatureStatus status = g_mime_signature_get_status (signature);
|
|
printf ("\"status\": %s",
|
|
json_quote_str (ctx_quote,
|
|
signature_status_to_string (status)));
|
|
|
|
GMimeCertificate *certificate = g_mime_signature_get_certificate (signature);
|
|
if (status == GMIME_SIGNATURE_STATUS_GOOD) {
|
|
if (certificate)
|
|
printf (", \"fingerprint\": %s", json_quote_str (ctx_quote, g_mime_certificate_get_fingerprint (certificate)));
|
|
/* these dates are seconds since the epoch; should we
|
|
* provide a more human-readable format string? */
|
|
time_t created = g_mime_signature_get_created (signature);
|
|
if (created != -1)
|
|
printf (", \"created\": %d", (int) created);
|
|
time_t expires = g_mime_signature_get_expires (signature);
|
|
if (expires > 0)
|
|
printf (", \"expires\": %d", (int) expires);
|
|
/* output user id only if validity is FULL or ULTIMATE. */
|
|
/* note that gmime is using the term "trust" here, which
|
|
* is WRONG. It's actually user id "validity". */
|
|
if (certificate) {
|
|
const char *name = g_mime_certificate_get_name (certificate);
|
|
GMimeCertificateTrust trust = g_mime_certificate_get_trust (certificate);
|
|
if (name && (trust == GMIME_CERTIFICATE_TRUST_FULLY || trust == GMIME_CERTIFICATE_TRUST_ULTIMATE))
|
|
printf (", \"userid\": %s", json_quote_str (ctx_quote, name));
|
|
}
|
|
} else if (certificate) {
|
|
const char *key_id = g_mime_certificate_get_key_id (certificate);
|
|
if (key_id)
|
|
printf (", \"keyid\": %s", json_quote_str (ctx_quote, key_id));
|
|
}
|
|
|
|
GMimeSignatureError errors = g_mime_signature_get_errors (signature);
|
|
if (errors != GMIME_SIGNATURE_ERROR_NONE) {
|
|
printf (", \"errors\": %d", errors);
|
|
}
|
|
|
|
printf ("}");
|
|
}
|
|
|
|
printf ("]");
|
|
|
|
talloc_free (ctx_quote);
|
|
}
|
|
#else
|
|
static void
|
|
format_part_sigstatus_json (mime_node_t *node)
|
|
{
|
|
const GMimeSignatureValidity* validity = node->sig_validity;
|
|
|
|
printf ("[");
|
|
|
|
if (!validity) {
|
|
printf ("]");
|
|
return;
|
|
}
|
|
|
|
const GMimeSigner *signer = g_mime_signature_validity_get_signers (validity);
|
|
int first = 1;
|
|
void *ctx_quote = talloc_new (NULL);
|
|
|
|
while (signer) {
|
|
if (first)
|
|
first = 0;
|
|
else
|
|
printf (", ");
|
|
|
|
printf ("{");
|
|
|
|
/* status */
|
|
printf ("\"status\": %s",
|
|
json_quote_str (ctx_quote,
|
|
signer_status_to_string (signer->status)));
|
|
|
|
if (signer->status == GMIME_SIGNER_STATUS_GOOD)
|
|
{
|
|
if (signer->fingerprint)
|
|
printf (", \"fingerprint\": %s", json_quote_str (ctx_quote, signer->fingerprint));
|
|
/* these dates are seconds since the epoch; should we
|
|
* provide a more human-readable format string? */
|
|
if (signer->created)
|
|
printf (", \"created\": %d", (int) signer->created);
|
|
if (signer->expires)
|
|
printf (", \"expires\": %d", (int) signer->expires);
|
|
/* output user id only if validity is FULL or ULTIMATE. */
|
|
/* note that gmime is using the term "trust" here, which
|
|
* is WRONG. It's actually user id "validity". */
|
|
if ((signer->name) && (signer->trust)) {
|
|
if ((signer->trust == GMIME_SIGNER_TRUST_FULLY) || (signer->trust == GMIME_SIGNER_TRUST_ULTIMATE))
|
|
printf (", \"userid\": %s", json_quote_str (ctx_quote, signer->name));
|
|
}
|
|
} else {
|
|
if (signer->keyid)
|
|
printf (", \"keyid\": %s", json_quote_str (ctx_quote, signer->keyid));
|
|
}
|
|
if (signer->errors != GMIME_SIGNER_ERROR_NONE) {
|
|
printf (", \"errors\": %d", signer->errors);
|
|
}
|
|
|
|
printf ("}");
|
|
signer = signer->next;
|
|
}
|
|
|
|
printf ("]");
|
|
|
|
talloc_free (ctx_quote);
|
|
}
|
|
#endif
|
|
|
|
static notmuch_status_t
|
|
format_part_text (const void *ctx, mime_node_t *node,
|
|
int indent, const notmuch_show_params_t *params)
|
|
{
|
|
/* The disposition and content-type metadata are associated with
|
|
* the envelope for message parts */
|
|
GMimeObject *meta = node->envelope_part ?
|
|
GMIME_OBJECT (node->envelope_part) : node->part;
|
|
GMimeContentType *content_type = g_mime_object_get_content_type (meta);
|
|
const notmuch_bool_t leaf = GMIME_IS_PART (node->part);
|
|
const char *part_type;
|
|
int i;
|
|
|
|
if (node->envelope_file) {
|
|
notmuch_message_t *message = node->envelope_file;
|
|
|
|
part_type = "message";
|
|
printf ("\f%s{ id:%s depth:%d match:%d excluded:%d filename:%s\n",
|
|
part_type,
|
|
notmuch_message_get_message_id (message),
|
|
indent,
|
|
notmuch_message_get_flag (message, NOTMUCH_MESSAGE_FLAG_MATCH) ? 1 : 0,
|
|
notmuch_message_get_flag (message, NOTMUCH_MESSAGE_FLAG_EXCLUDED) ? 1 : 0,
|
|
notmuch_message_get_filename (message));
|
|
} else {
|
|
GMimeContentDisposition *disposition = g_mime_object_get_content_disposition (meta);
|
|
const char *cid = g_mime_object_get_content_id (meta);
|
|
const char *filename = leaf ?
|
|
g_mime_part_get_filename (GMIME_PART (node->part)) : NULL;
|
|
|
|
if (disposition &&
|
|
strcmp (disposition->disposition, GMIME_DISPOSITION_ATTACHMENT) == 0)
|
|
part_type = "attachment";
|
|
else
|
|
part_type = "part";
|
|
|
|
printf ("\f%s{ ID: %d", part_type, node->part_num);
|
|
if (filename)
|
|
printf (", Filename: %s", filename);
|
|
if (cid)
|
|
printf (", Content-id: %s", cid);
|
|
printf (", Content-type: %s\n", g_mime_content_type_to_string (content_type));
|
|
}
|
|
|
|
if (GMIME_IS_MESSAGE (node->part)) {
|
|
GMimeMessage *message = GMIME_MESSAGE (node->part);
|
|
InternetAddressList *recipients;
|
|
const char *recipients_string;
|
|
|
|
printf ("\fheader{\n");
|
|
if (node->envelope_file)
|
|
printf ("%s\n", _get_one_line_summary (ctx, node->envelope_file));
|
|
printf ("Subject: %s\n", g_mime_message_get_subject (message));
|
|
printf ("From: %s\n", g_mime_message_get_sender (message));
|
|
recipients = g_mime_message_get_recipients (message, GMIME_RECIPIENT_TYPE_TO);
|
|
recipients_string = internet_address_list_to_string (recipients, 0);
|
|
if (recipients_string)
|
|
printf ("To: %s\n", recipients_string);
|
|
recipients = g_mime_message_get_recipients (message, GMIME_RECIPIENT_TYPE_CC);
|
|
recipients_string = internet_address_list_to_string (recipients, 0);
|
|
if (recipients_string)
|
|
printf ("Cc: %s\n", recipients_string);
|
|
printf ("Date: %s\n", g_mime_message_get_date_as_string (message));
|
|
printf ("\fheader}\n");
|
|
|
|
printf ("\fbody{\n");
|
|
}
|
|
|
|
if (leaf) {
|
|
if (g_mime_content_type_is_type (content_type, "text", "*") &&
|
|
!g_mime_content_type_is_type (content_type, "text", "html"))
|
|
{
|
|
GMimeStream *stream_stdout = g_mime_stream_file_new (stdout);
|
|
g_mime_stream_file_set_owner (GMIME_STREAM_FILE (stream_stdout), FALSE);
|
|
show_text_part_content (node->part, stream_stdout, 0);
|
|
g_object_unref(stream_stdout);
|
|
} else {
|
|
printf ("Non-text part: %s\n",
|
|
g_mime_content_type_to_string (content_type));
|
|
}
|
|
}
|
|
|
|
for (i = 0; i < node->nchildren; i++)
|
|
format_part_text (ctx, mime_node_child (node, i), indent, params);
|
|
|
|
if (GMIME_IS_MESSAGE (node->part))
|
|
printf ("\fbody}\n");
|
|
|
|
printf ("\f%s}\n", part_type);
|
|
|
|
return NOTMUCH_STATUS_SUCCESS;
|
|
}
|
|
|
|
void
|
|
format_part_json (const void *ctx, mime_node_t *node, notmuch_bool_t first, notmuch_bool_t output_body)
|
|
{
|
|
/* Any changes to the JSON format should be reflected in the file
|
|
* devel/schemata. */
|
|
|
|
if (node->envelope_file) {
|
|
printf ("{");
|
|
format_message_json (ctx, node->envelope_file);
|
|
|
|
printf ("\"headers\": ");
|
|
format_headers_json (ctx, GMIME_MESSAGE (node->part), FALSE);
|
|
|
|
if (output_body) {
|
|
printf (", \"body\": [");
|
|
format_part_json (ctx, mime_node_child (node, 0), first, TRUE);
|
|
printf ("]");
|
|
}
|
|
printf ("}");
|
|
return;
|
|
}
|
|
|
|
void *local = talloc_new (ctx);
|
|
/* The disposition and content-type metadata are associated with
|
|
* the envelope for message parts */
|
|
GMimeObject *meta = node->envelope_part ?
|
|
GMIME_OBJECT (node->envelope_part) : node->part;
|
|
GMimeContentType *content_type = g_mime_object_get_content_type (meta);
|
|
const char *cid = g_mime_object_get_content_id (meta);
|
|
const char *filename = GMIME_IS_PART (node->part) ?
|
|
g_mime_part_get_filename (GMIME_PART (node->part)) : NULL;
|
|
const char *terminator = "";
|
|
int i;
|
|
|
|
if (!first)
|
|
printf (", ");
|
|
|
|
printf ("{\"id\": %d", node->part_num);
|
|
|
|
if (node->decrypt_attempted)
|
|
printf (", \"encstatus\": [{\"status\": \"%s\"}]",
|
|
node->decrypt_success ? "good" : "bad");
|
|
|
|
if (node->verify_attempted) {
|
|
printf (", \"sigstatus\": ");
|
|
format_part_sigstatus_json (node);
|
|
}
|
|
|
|
printf (", \"content-type\": %s",
|
|
json_quote_str (local, g_mime_content_type_to_string (content_type)));
|
|
|
|
if (cid)
|
|
printf (", \"content-id\": %s", json_quote_str (local, cid));
|
|
|
|
if (filename)
|
|
printf (", \"filename\": %s", json_quote_str (local, filename));
|
|
|
|
if (GMIME_IS_PART (node->part)) {
|
|
/* For non-HTML text parts, we include the content in the
|
|
* JSON. Since JSON must be Unicode, we handle charset
|
|
* decoding here and do not report a charset to the caller.
|
|
* For text/html parts, we do not include the content. If a
|
|
* caller is interested in text/html parts, it should retrieve
|
|
* them separately and they will not be decoded. Since this
|
|
* makes charset decoding the responsibility on the caller, we
|
|
* report the charset for text/html parts.
|
|
*/
|
|
if (g_mime_content_type_is_type (content_type, "text", "html")) {
|
|
const char *content_charset = g_mime_object_get_content_type_parameter (meta, "charset");
|
|
|
|
if (content_charset != NULL)
|
|
printf (", \"content-charset\": %s", json_quote_str (local, content_charset));
|
|
} else if (g_mime_content_type_is_type (content_type, "text", "*")) {
|
|
GMimeStream *stream_memory = g_mime_stream_mem_new ();
|
|
GByteArray *part_content;
|
|
show_text_part_content (node->part, stream_memory, 0);
|
|
part_content = g_mime_stream_mem_get_byte_array (GMIME_STREAM_MEM (stream_memory));
|
|
|
|
printf (", \"content\": %s", json_quote_chararray (local, (char *) part_content->data, part_content->len));
|
|
g_object_unref (stream_memory);
|
|
}
|
|
} else if (GMIME_IS_MULTIPART (node->part)) {
|
|
printf (", \"content\": [");
|
|
terminator = "]";
|
|
} else if (GMIME_IS_MESSAGE (node->part)) {
|
|
printf (", \"content\": [{");
|
|
printf ("\"headers\": ");
|
|
format_headers_json (local, GMIME_MESSAGE (node->part), FALSE);
|
|
|
|
printf (", \"body\": [");
|
|
terminator = "]}]";
|
|
}
|
|
|
|
talloc_free (local);
|
|
|
|
for (i = 0; i < node->nchildren; i++)
|
|
format_part_json (ctx, mime_node_child (node, i), i == 0, TRUE);
|
|
|
|
printf ("%s}", terminator);
|
|
}
|
|
|
|
static notmuch_status_t
|
|
format_part_json_entry (const void *ctx, mime_node_t *node, unused (int indent),
|
|
const notmuch_show_params_t *params)
|
|
{
|
|
format_part_json (ctx, node, TRUE, params->output_body);
|
|
|
|
return NOTMUCH_STATUS_SUCCESS;
|
|
}
|
|
|
|
/* Print a message in "mboxrd" format as documented, for example,
|
|
* here:
|
|
*
|
|
* http://qmail.org/qmail-manual-html/man5/mbox.html
|
|
*/
|
|
static notmuch_status_t
|
|
format_part_mbox (const void *ctx, mime_node_t *node, unused (int indent),
|
|
unused (const notmuch_show_params_t *params))
|
|
{
|
|
notmuch_message_t *message = node->envelope_file;
|
|
|
|
const char *filename;
|
|
FILE *file;
|
|
const char *from;
|
|
|
|
time_t date;
|
|
struct tm date_gmtime;
|
|
char date_asctime[26];
|
|
|
|
char *line = NULL;
|
|
size_t line_size;
|
|
ssize_t line_len;
|
|
|
|
if (!message)
|
|
INTERNAL_ERROR ("format_part_mbox requires a root part");
|
|
|
|
filename = notmuch_message_get_filename (message);
|
|
file = fopen (filename, "r");
|
|
if (file == NULL) {
|
|
fprintf (stderr, "Failed to open %s: %s\n",
|
|
filename, strerror (errno));
|
|
return NOTMUCH_STATUS_FILE_ERROR;
|
|
}
|
|
|
|
from = notmuch_message_get_header (message, "from");
|
|
from = _extract_email_address (ctx, from);
|
|
|
|
date = notmuch_message_get_date (message);
|
|
gmtime_r (&date, &date_gmtime);
|
|
asctime_r (&date_gmtime, date_asctime);
|
|
|
|
printf ("From %s %s", from, date_asctime);
|
|
|
|
while ((line_len = getline (&line, &line_size, file)) != -1 ) {
|
|
if (_is_from_line (line))
|
|
putchar ('>');
|
|
printf ("%s", line);
|
|
}
|
|
|
|
printf ("\n");
|
|
|
|
fclose (file);
|
|
|
|
return NOTMUCH_STATUS_SUCCESS;
|
|
}
|
|
|
|
static notmuch_status_t
|
|
format_part_raw (unused (const void *ctx), mime_node_t *node,
|
|
unused (int indent),
|
|
unused (const notmuch_show_params_t *params))
|
|
{
|
|
if (node->envelope_file) {
|
|
/* Special case the entire message to avoid MIME parsing. */
|
|
const char *filename;
|
|
FILE *file;
|
|
size_t size;
|
|
char buf[4096];
|
|
|
|
filename = notmuch_message_get_filename (node->envelope_file);
|
|
if (filename == NULL) {
|
|
fprintf (stderr, "Error: Cannot get message filename.\n");
|
|
return NOTMUCH_STATUS_FILE_ERROR;
|
|
}
|
|
|
|
file = fopen (filename, "r");
|
|
if (file == NULL) {
|
|
fprintf (stderr, "Error: Cannot open file %s: %s\n", filename, strerror (errno));
|
|
return NOTMUCH_STATUS_FILE_ERROR;
|
|
}
|
|
|
|
while (!feof (file)) {
|
|
size = fread (buf, 1, sizeof (buf), file);
|
|
if (ferror (file)) {
|
|
fprintf (stderr, "Error: Read failed from %s\n", filename);
|
|
fclose (file);
|
|
return NOTMUCH_STATUS_FILE_ERROR;
|
|
}
|
|
|
|
if (fwrite (buf, size, 1, stdout) != 1) {
|
|
fprintf (stderr, "Error: Write failed\n");
|
|
fclose (file);
|
|
return NOTMUCH_STATUS_FILE_ERROR;
|
|
}
|
|
}
|
|
|
|
fclose (file);
|
|
return NOTMUCH_STATUS_SUCCESS;
|
|
}
|
|
|
|
GMimeStream *stream_stdout;
|
|
GMimeStream *stream_filter = NULL;
|
|
|
|
stream_stdout = g_mime_stream_file_new (stdout);
|
|
g_mime_stream_file_set_owner (GMIME_STREAM_FILE (stream_stdout), FALSE);
|
|
|
|
stream_filter = g_mime_stream_filter_new (stream_stdout);
|
|
|
|
if (GMIME_IS_PART (node->part)) {
|
|
/* For leaf parts, we emit only the transfer-decoded
|
|
* body. */
|
|
GMimeDataWrapper *wrapper;
|
|
wrapper = g_mime_part_get_content_object (GMIME_PART (node->part));
|
|
|
|
if (wrapper && stream_filter)
|
|
g_mime_data_wrapper_write_to_stream (wrapper, stream_filter);
|
|
} else {
|
|
/* Write out the whole part. For message parts (the root
|
|
* part and embedded message parts), this will be the
|
|
* message including its headers (but not the
|
|
* encapsulating part's headers). For multipart parts,
|
|
* this will include the headers. */
|
|
if (stream_filter)
|
|
g_mime_object_write_to_stream (node->part, stream_filter);
|
|
}
|
|
|
|
if (stream_filter)
|
|
g_object_unref (stream_filter);
|
|
|
|
if (stream_stdout)
|
|
g_object_unref(stream_stdout);
|
|
|
|
return NOTMUCH_STATUS_SUCCESS;
|
|
}
|
|
|
|
static notmuch_status_t
|
|
show_null_message (const notmuch_show_format_t *format)
|
|
{
|
|
/* Output a null message. Currently empty for all formats except Json */
|
|
if (format->null_message)
|
|
printf ("%s", format->null_message);
|
|
return NOTMUCH_STATUS_SUCCESS;
|
|
}
|
|
|
|
static notmuch_status_t
|
|
show_message (void *ctx,
|
|
const notmuch_show_format_t *format,
|
|
notmuch_message_t *message,
|
|
int indent,
|
|
notmuch_show_params_t *params)
|
|
{
|
|
void *local = talloc_new (ctx);
|
|
mime_node_t *root, *part;
|
|
notmuch_status_t status;
|
|
|
|
status = mime_node_open (local, message, &(params->crypto), &root);
|
|
if (status)
|
|
goto DONE;
|
|
part = mime_node_seek_dfs (root, (params->part < 0 ? 0 : params->part));
|
|
if (part)
|
|
status = format->part (local, part, indent, params);
|
|
DONE:
|
|
talloc_free (local);
|
|
return status;
|
|
}
|
|
|
|
static notmuch_status_t
|
|
show_messages (void *ctx,
|
|
const notmuch_show_format_t *format,
|
|
notmuch_messages_t *messages,
|
|
int indent,
|
|
notmuch_show_params_t *params)
|
|
{
|
|
notmuch_message_t *message;
|
|
notmuch_bool_t match;
|
|
notmuch_bool_t excluded;
|
|
int first_set = 1;
|
|
int next_indent;
|
|
notmuch_status_t status, res = NOTMUCH_STATUS_SUCCESS;
|
|
|
|
if (format->message_set_start)
|
|
fputs (format->message_set_start, stdout);
|
|
|
|
for (;
|
|
notmuch_messages_valid (messages);
|
|
notmuch_messages_move_to_next (messages))
|
|
{
|
|
if (!first_set && format->message_set_sep)
|
|
fputs (format->message_set_sep, stdout);
|
|
first_set = 0;
|
|
|
|
if (format->message_set_start)
|
|
fputs (format->message_set_start, stdout);
|
|
|
|
message = notmuch_messages_get (messages);
|
|
|
|
match = notmuch_message_get_flag (message, NOTMUCH_MESSAGE_FLAG_MATCH);
|
|
excluded = notmuch_message_get_flag (message, NOTMUCH_MESSAGE_FLAG_EXCLUDED);
|
|
|
|
next_indent = indent;
|
|
|
|
if ((match && (!excluded || !params->omit_excluded)) || params->entire_thread) {
|
|
status = show_message (ctx, format, message, indent, params);
|
|
if (status && !res)
|
|
res = status;
|
|
next_indent = indent + 1;
|
|
} else {
|
|
status = show_null_message (format);
|
|
}
|
|
|
|
if (!status && format->message_set_sep)
|
|
fputs (format->message_set_sep, stdout);
|
|
|
|
status = show_messages (ctx,
|
|
format,
|
|
notmuch_message_get_replies (message),
|
|
next_indent,
|
|
params);
|
|
if (status && !res)
|
|
res = status;
|
|
|
|
notmuch_message_destroy (message);
|
|
|
|
if (format->message_set_end)
|
|
fputs (format->message_set_end, stdout);
|
|
}
|
|
|
|
if (format->message_set_end)
|
|
fputs (format->message_set_end, stdout);
|
|
|
|
return res;
|
|
}
|
|
|
|
/* Formatted output of single message */
|
|
static int
|
|
do_show_single (void *ctx,
|
|
notmuch_query_t *query,
|
|
const notmuch_show_format_t *format,
|
|
notmuch_show_params_t *params)
|
|
{
|
|
notmuch_messages_t *messages;
|
|
notmuch_message_t *message;
|
|
|
|
if (notmuch_query_count_messages (query) != 1) {
|
|
fprintf (stderr, "Error: search term did not match precisely one message.\n");
|
|
return 1;
|
|
}
|
|
|
|
messages = notmuch_query_search_messages (query);
|
|
message = notmuch_messages_get (messages);
|
|
|
|
if (message == NULL) {
|
|
fprintf (stderr, "Error: Cannot find matching message.\n");
|
|
return 1;
|
|
}
|
|
|
|
notmuch_message_set_flag (message, NOTMUCH_MESSAGE_FLAG_MATCH, 1);
|
|
|
|
return show_message (ctx, format, message, 0, params) != NOTMUCH_STATUS_SUCCESS;
|
|
}
|
|
|
|
/* Formatted output of threads */
|
|
static int
|
|
do_show (void *ctx,
|
|
notmuch_query_t *query,
|
|
const notmuch_show_format_t *format,
|
|
notmuch_show_params_t *params)
|
|
{
|
|
notmuch_threads_t *threads;
|
|
notmuch_thread_t *thread;
|
|
notmuch_messages_t *messages;
|
|
int first_toplevel = 1;
|
|
notmuch_status_t status, res = NOTMUCH_STATUS_SUCCESS;
|
|
|
|
if (format->message_set_start)
|
|
fputs (format->message_set_start, stdout);
|
|
|
|
for (threads = notmuch_query_search_threads (query);
|
|
notmuch_threads_valid (threads);
|
|
notmuch_threads_move_to_next (threads))
|
|
{
|
|
thread = notmuch_threads_get (threads);
|
|
|
|
messages = notmuch_thread_get_toplevel_messages (thread);
|
|
|
|
if (messages == NULL)
|
|
INTERNAL_ERROR ("Thread %s has no toplevel messages.\n",
|
|
notmuch_thread_get_thread_id (thread));
|
|
|
|
if (!first_toplevel && format->message_set_sep)
|
|
fputs (format->message_set_sep, stdout);
|
|
first_toplevel = 0;
|
|
|
|
status = show_messages (ctx, format, messages, 0, params);
|
|
if (status && !res)
|
|
res = status;
|
|
|
|
notmuch_thread_destroy (thread);
|
|
|
|
}
|
|
|
|
if (format->message_set_end)
|
|
fputs (format->message_set_end, stdout);
|
|
|
|
return res != NOTMUCH_STATUS_SUCCESS;
|
|
}
|
|
|
|
enum {
|
|
NOTMUCH_FORMAT_NOT_SPECIFIED,
|
|
NOTMUCH_FORMAT_JSON,
|
|
NOTMUCH_FORMAT_TEXT,
|
|
NOTMUCH_FORMAT_MBOX,
|
|
NOTMUCH_FORMAT_RAW
|
|
};
|
|
|
|
enum {
|
|
ENTIRE_THREAD_DEFAULT,
|
|
ENTIRE_THREAD_TRUE,
|
|
ENTIRE_THREAD_FALSE,
|
|
};
|
|
|
|
/* The following is to allow future options to be added more easily */
|
|
enum {
|
|
EXCLUDE_TRUE,
|
|
EXCLUDE_FALSE,
|
|
};
|
|
|
|
int
|
|
notmuch_show_command (void *ctx, unused (int argc), unused (char *argv[]))
|
|
{
|
|
notmuch_config_t *config;
|
|
notmuch_database_t *notmuch;
|
|
notmuch_query_t *query;
|
|
char *query_string;
|
|
int opt_index, ret;
|
|
const notmuch_show_format_t *format = &format_text;
|
|
notmuch_show_params_t params = {
|
|
.part = -1,
|
|
.omit_excluded = TRUE,
|
|
.output_body = TRUE,
|
|
.crypto = {
|
|
.verify = FALSE,
|
|
.decrypt = FALSE
|
|
}
|
|
};
|
|
int format_sel = NOTMUCH_FORMAT_NOT_SPECIFIED;
|
|
int exclude = EXCLUDE_TRUE;
|
|
int entire_thread = ENTIRE_THREAD_DEFAULT;
|
|
|
|
notmuch_opt_desc_t options[] = {
|
|
{ NOTMUCH_OPT_KEYWORD, &format_sel, "format", 'f',
|
|
(notmuch_keyword_t []){ { "json", NOTMUCH_FORMAT_JSON },
|
|
{ "text", NOTMUCH_FORMAT_TEXT },
|
|
{ "mbox", NOTMUCH_FORMAT_MBOX },
|
|
{ "raw", NOTMUCH_FORMAT_RAW },
|
|
{ 0, 0 } } },
|
|
{ NOTMUCH_OPT_KEYWORD, &exclude, "exclude", 'x',
|
|
(notmuch_keyword_t []){ { "true", EXCLUDE_TRUE },
|
|
{ "false", EXCLUDE_FALSE },
|
|
{ 0, 0 } } },
|
|
{ NOTMUCH_OPT_KEYWORD, &entire_thread, "entire-thread", 't',
|
|
(notmuch_keyword_t []){ { "true", ENTIRE_THREAD_TRUE },
|
|
{ "false", ENTIRE_THREAD_FALSE },
|
|
{ "", ENTIRE_THREAD_TRUE },
|
|
{ 0, 0 } } },
|
|
{ NOTMUCH_OPT_INT, ¶ms.part, "part", 'p', 0 },
|
|
{ NOTMUCH_OPT_BOOLEAN, ¶ms.crypto.decrypt, "decrypt", 'd', 0 },
|
|
{ NOTMUCH_OPT_BOOLEAN, ¶ms.crypto.verify, "verify", 'v', 0 },
|
|
{ NOTMUCH_OPT_BOOLEAN, ¶ms.output_body, "body", 'b', 0 },
|
|
{ 0, 0, 0, 0, 0 }
|
|
};
|
|
|
|
opt_index = parse_arguments (argc, argv, options, 1);
|
|
if (opt_index < 0) {
|
|
/* diagnostics already printed */
|
|
return 1;
|
|
}
|
|
|
|
/* decryption implies verification */
|
|
if (params.crypto.decrypt)
|
|
params.crypto.verify = TRUE;
|
|
|
|
if (format_sel == NOTMUCH_FORMAT_NOT_SPECIFIED) {
|
|
/* if part was requested and format was not specified, use format=raw */
|
|
if (params.part >= 0)
|
|
format_sel = NOTMUCH_FORMAT_RAW;
|
|
else
|
|
format_sel = NOTMUCH_FORMAT_TEXT;
|
|
}
|
|
|
|
switch (format_sel) {
|
|
case NOTMUCH_FORMAT_JSON:
|
|
format = &format_json;
|
|
break;
|
|
case NOTMUCH_FORMAT_TEXT:
|
|
format = &format_text;
|
|
break;
|
|
case NOTMUCH_FORMAT_MBOX:
|
|
if (params.part > 0) {
|
|
fprintf (stderr, "Error: specifying parts is incompatible with mbox output format.\n");
|
|
return 1;
|
|
}
|
|
|
|
format = &format_mbox;
|
|
break;
|
|
case NOTMUCH_FORMAT_RAW:
|
|
format = &format_raw;
|
|
/* If --format=raw specified without specifying part, we can only
|
|
* output single message, so set part=0 */
|
|
if (params.part < 0)
|
|
params.part = 0;
|
|
params.raw = TRUE;
|
|
break;
|
|
}
|
|
|
|
/* Default is entire-thread = FALSE except for format=json. */
|
|
if (entire_thread == ENTIRE_THREAD_DEFAULT) {
|
|
if (format == &format_json)
|
|
entire_thread = ENTIRE_THREAD_TRUE;
|
|
else
|
|
entire_thread = ENTIRE_THREAD_FALSE;
|
|
}
|
|
|
|
if (!params.output_body) {
|
|
if (params.part > 0) {
|
|
fprintf (stderr, "Warning: --body=false is incompatible with --part > 0. Disabling.\n");
|
|
params.output_body = TRUE;
|
|
} else {
|
|
if (format != &format_json)
|
|
fprintf (stderr, "Warning: --body=false only implemented for format=json\n");
|
|
}
|
|
}
|
|
|
|
if (entire_thread == ENTIRE_THREAD_TRUE)
|
|
params.entire_thread = TRUE;
|
|
else
|
|
params.entire_thread = FALSE;
|
|
|
|
config = notmuch_config_open (ctx, NULL, NULL);
|
|
if (config == NULL)
|
|
return 1;
|
|
|
|
query_string = query_string_from_args (ctx, argc-opt_index, argv+opt_index);
|
|
if (query_string == NULL) {
|
|
fprintf (stderr, "Out of memory\n");
|
|
return 1;
|
|
}
|
|
|
|
if (*query_string == '\0') {
|
|
fprintf (stderr, "Error: notmuch show requires at least one search term.\n");
|
|
return 1;
|
|
}
|
|
|
|
if (notmuch_database_open (notmuch_config_get_database_path (config),
|
|
NOTMUCH_DATABASE_MODE_READ_ONLY, ¬much))
|
|
return 1;
|
|
|
|
query = notmuch_query_create (notmuch, query_string);
|
|
if (query == NULL) {
|
|
fprintf (stderr, "Out of memory\n");
|
|
return 1;
|
|
}
|
|
|
|
/* If a single message is requested we do not use search_excludes. */
|
|
if (params.part >= 0)
|
|
ret = do_show_single (ctx, query, format, ¶ms);
|
|
else {
|
|
/* We always apply set the exclude flag. The
|
|
* exclude=true|false option controls whether or not we return
|
|
* threads that only match in an excluded message */
|
|
const char **search_exclude_tags;
|
|
size_t search_exclude_tags_length;
|
|
unsigned int i;
|
|
|
|
search_exclude_tags = notmuch_config_get_search_exclude_tags
|
|
(config, &search_exclude_tags_length);
|
|
for (i = 0; i < search_exclude_tags_length; i++)
|
|
notmuch_query_add_tag_exclude (query, search_exclude_tags[i]);
|
|
|
|
if (exclude == EXCLUDE_FALSE) {
|
|
notmuch_query_set_omit_excluded (query, FALSE);
|
|
params.omit_excluded = FALSE;
|
|
}
|
|
|
|
ret = do_show (ctx, query, format, ¶ms);
|
|
}
|
|
|
|
notmuch_crypto_cleanup (¶ms.crypto);
|
|
notmuch_query_destroy (query);
|
|
notmuch_database_destroy (notmuch);
|
|
|
|
return ret;
|
|
}
|