Add support for hash chaining to detect modifications in postings

doc/ledger.1

@@ -1,4 +1,4 @@
-.Dd March 15, 2019
+.Dd November 30, 2023
 .Dt LEDGER 1
 .Os
 .Sh NAME
@@ -634,6 +634,24 @@ Set the format for the headers that separate reports section of
 a grouped report.  Only has effect with a
 .Fl \-group-by Ar EXPR
 register report.
+.It Fl \-hashes Ar ALGO
+Record the hash of each transaction in a
+.Ar Hash
+metadata value, according to the hashing algorithm given by the
+.Ar ALGO
+argument. Support algorithms are
+.Ar sha512
+or
+.Ar sh512_256
+where both use SHA512, but the latter only stores the first half of
+the hash value.
+.Pp
+If a
+.Ar Hash
+metadata value is explicitly provided and does not match what would
+have been generated, an error is reported. Hashes depend on previous
+entries, such that setting a single hash value is sufficient to
+guarantee the shape of the entire history leading up to that entry.
 .It Fl \-head Ar INT
 Print the first
 .Ar INT

doc/ledger3.texi

@@ -6697,6 +6697,57 @@ $ ledger reg Expenses --group-by "payee" --group-title-format "-----------------
 ...
 @end smallexample
 
+@item --hashes @var{ALGO}
+Records the chained hash of each transaction in a @var{Hash} metadata
+value, according to the hashing algorithm given by the @var{ALGO}
+argument (at the moment, only @code{sha512} is supported). To use this,
+record the @var{Hash} metadata explicitly in some of your transactions;
+these will be checked against the hashes calculated internally, and if
+they do not match, an error is reported. You may also write just a
+prefix of the @var{Hash}, which is less verbose but still gives quite
+good assurance.
+
+The support algorithms are:
+
+@table @code
+@item sha512
+Use the SHA512 hashing algorithm.
+@item sha512_256
+Same as SHA512, but record only the first 256 bits.
+@end table
+
+Somewhat like balance assertions, which give assurance that previous
+posting amounts are correct, these @var{Hash} tags give assurance that
+all previous journal entries (in parse order) are unchanged (or at
+least, their combined hash matches the Hash tag currently appearing in
+the journal).
+
+These hashes depend on the hashes of previous transactions, such that
+the single hash value of the final transaction is sufficient to
+guarantee the shape of the entire history leading up to it.
+
+The other details that the hash depends on are the following details
+from each posting in the transaction:
+
+@itemize
+@item fullname of the account
+@item amount value
+@end itemize
+
+In addition, these details are hashed from the transaction itself:
+
+@itemize
+@item actual date
+@item auxiliary date (if provided)
+@item code (if provided)
+@item payee
+@end itemize
+
+This list also means that changes in the comments of postings or
+transactions, or in the ordering of the postings within a transaction,
+will not affect the hash. The ordering of the transactions does matter,
+however, the same way as it does for balance assertions.
+
 @item --head @var{INT}
 @itemx --first @var{INT}
 Print the first @var{INT} entries.  Opposite of @option{--tail

src/CMakeLists.txt

@@ -51,7 +51,8 @@ set(LEDGER_SOURCES
   times.cc
   error.cc
   utils.cc
-  wcwidth.cc)
+  wcwidth.cc
+  sha512.cc)
 
 if (HAVE_GPGME)
   list(APPEND LEDGER_SOURCES

src/generate.cc

@@ -365,7 +365,7 @@ void generate_posts_iterator::increment()
       parsing_context.get_current().journal = session.journal.get();
       parsing_context.get_current().scope   = &session;
 
-      if (session.journal->read(parsing_context) != 0) {
+      if (session.journal->read(parsing_context, NO_HASHES) != 0) {
         VERIFY(session.journal->xacts.back()->valid());
         posts.reset(*session.journal->xacts.back());
         post = *posts++;

src/global.cc

@@ -116,7 +116,7 @@ void global_scope_t::parse_init(path init_file)
   parsing_context.get_current().journal = session().journal.get();
   parsing_context.get_current().scope   = &report();
 
-  if (session().journal->read(parsing_context) > 0 ||
+  if (session().journal->read(parsing_context, NO_HASHES) > 0 ||
       session().journal->auto_xacts.size() > 0 ||
       session().journal->period_xacts.size() > 0) {
     throw_(parse_error, _f("Transactions found in initialization file '%1%'")

src/item.cc

@@ -133,13 +133,17 @@ item_t::set_tag(const string&            tag,
 
   string_map::iterator i = metadata->find(tag);
   if (i == metadata->end()) {
+    DEBUG("item.meta", "Setting new metadata value");
     std::pair<string_map::iterator, bool> result
       = metadata->insert(string_map::value_type(tag, tag_data_t(data, false)));
     assert(result.second);
     return result.first;
   } else {
-    if (overwrite_existing)
+    DEBUG("item.meta", "Found old metadata value");
+    if (overwrite_existing) {
+      DEBUG("item.meta", "Overwriting old metadata value");
       (*i).second = tag_data_t(data, false);
+    }
     return i;
   }
 }

src/journal.cc

@@ -466,7 +466,8 @@ bool journal_t::remove_xact(xact_t * xact)
   return true;
 }
 
-std::size_t journal_t::read(parse_context_stack_t& context)
+std::size_t journal_t::read(parse_context_stack_t& context,
+                            hash_type_t hash_type)
 {
   std::size_t count = 0;
   try {
@@ -485,7 +486,7 @@ std::size_t journal_t::read(parse_context_stack_t& context)
     if (! current.master)
       current.master = master;
 
-    count = read_textual(context);
+    count = read_textual(context, hash_type);
     if (count > 0) {
       if (! current.pathname.empty())
         sources.push_back(fileinfo_t(current.pathname));

src/journal.h

@@ -184,7 +184,8 @@ class journal_t : public noncopyable
     return period_xacts.end();
   }
 
-  std::size_t read(parse_context_stack_t& context);
+  std::size_t read(parse_context_stack_t& context,
+                   hash_type_t hash_type);
 
   bool has_xdata();
   void clear_xdata();
@@ -193,7 +194,8 @@ class journal_t : public noncopyable
 
 private:
 
-  std::size_t read_textual(parse_context_stack_t& context);
+  std::size_t read_textual(parse_context_stack_t& context,
+                           hash_type_t hash_type);
 
   bool        should_check_payees();
   bool        payee_not_registered(const string& name);

src/precmd.cc

@@ -74,7 +74,7 @@ namespace {
         parsing_context.get_current().journal = report.session.journal.get();
         parsing_context.get_current().scope   = &report.session;
 
-        report.session.journal->read(parsing_context);
+        report.session.journal->read(parsing_context, NO_HASHES);
         report.session.journal->clear_xdata();
       }
     }

src/session.cc

@@ -134,7 +134,7 @@ std::size_t session_t::read_data(const string& master_account)
       parsing_context.push(*price_db_path);
       parsing_context.get_current().journal = journal.get();
       try {
-        if (journal->read(parsing_context) > 0)
+        if (journal->read(parsing_context, HANDLER(hashes_).hash_type) > 0)
           throw_(parse_error, _("Transactions not allowed in price history file"));
       }
       catch (...) {
@@ -169,7 +169,7 @@ std::size_t session_t::read_data(const string& master_account)
     parsing_context.get_current().journal = journal.get();
     parsing_context.get_current().master  = acct;
     try {
-      xact_count += journal->read(parsing_context);
+      xact_count += journal->read(parsing_context, HANDLER(hashes_).hash_type);
     }
     catch (...) {
       parsing_context.pop();
@@ -230,7 +230,7 @@ journal_t * session_t::read_journal_from_string(const string& data)
   parsing_context.get_current().journal = journal.get();
   parsing_context.get_current().master  = journal->master;
   try {
-    journal->read(parsing_context);
+    journal->read(parsing_context, HANDLER(hashes_).hash_type);
   }
   catch (...) {
     parsing_context.pop();
@@ -331,6 +331,9 @@ option_t<session_t> * session_t::lookup_option(const char * p)
   case 'f':
     OPT_(file_); // -f
     break;
+  case 'h':
+    OPT(hashes_);
+    break;
   case 'i':
     OPT(input_date_format_);
     break;

src/session.h

@@ -112,6 +112,7 @@ class session_t : public symbol_scope_t
     HANDLER(decimal_comma).report(out);
     HANDLER(time_colon).report(out);
     HANDLER(file_).report(out);
+    HANDLER(hashes_).report(out);
     HANDLER(input_date_format_).report(out);
     HANDLER(explicit).report(out);
     HANDLER(master_account_).report(out);
@@ -162,6 +163,20 @@ class session_t : public symbol_scope_t
      data_files.push_back(str);
    });
 
+  OPTION__
+  (session_t, hashes_,
+   hash_type_t hash_type = NO_HASHES;
+   CTOR(session_t, hashes_) {}
+   DO_(str) {
+     if (str == "sha512" || str == "SHA512") {
+       hash_type = HASH_SHA512;
+     } else if (str == "sha512_256" || str == "SHA512_256") {
+       hash_type = HASH_SHA512_256;
+     } else {
+        throw_(std::invalid_argument, _f("Unrecognized hash type"));
+     }
+   });
+
   OPTION_(session_t, input_date_format_, DO_(str) {
       // This changes static variables inside times.h, which affects the
       // basic date parser.