got references parsing

1 files changed, 192 insertions, 36 deletions

diff --git a/src/xref.c b/src/xref.c
index 6af8132..e1f3a9d 100644
--- a/src/xref.c
+++ b/src/xref.c
@@ -3,6 +3,7 @@
 #include <inttypes.h>
 #include <stdlib.h>
 #include <stdbool.h>
+#include <string.h>
 
 #define SLEN(x) (sizeof((x)) - 1)
 
@@ -14,7 +15,7 @@
 #define BC_DELM (1<<4) /* character is a delimiter */
 
 /**
- * byte classification
+ * pdf byte classification
  */
 uint8_t bclass[] = {
     BC_WSPC, BC_NONE, BC_NONE, BC_NONE, /* 00 - 03 */
@@ -126,7 +127,7 @@ struct cos_dictionary_entry {
 struct cos_reference {
     /** id of indirect object */
     uint64_t id;
-            
+
     /* generation of indirect object */
     uint64_t generation;
 };
@@ -136,19 +137,19 @@ struct cos_object {
     union {
         /** boolean */
         bool b;
-        
+
         /** integer */
         int64_t i;
-        
+
         /** real */
         double r;
-        
+
         /** name */
         char *n;
-        
+
         /** string */
         char *s;
-        
+
         /** stream data */
         uint8_t *stream;
 
@@ -157,7 +158,7 @@ struct cos_object {
 
         /** reference */
         struct cos_reference *reference;
-        
+
     } u;
 };
 
@@ -169,10 +170,10 @@ struct cos_indirect_object {
 
     /* reference identifier */
     struct cos_reference ref;
-    
+
     /** offset of object */
     uint64_t offset;
-    
+
     /* direct object */
     struct cos_object *o;
 };
@@ -367,7 +368,7 @@ int cos_indirect_object_add(struct pdf_doc *doc,
     nobj->offset = obj_offset;
 
     doc->cos_list = nobj;
-    
+
     printf("xref %"PRIu64" %"PRIu64" %"PRIu64"\n",
            obj_number, obj_offset, obj_generation);
     return 0;
@@ -390,10 +391,10 @@ int cos_free_object(struct cos_object *cos_obj)
         dentry = cos_obj->u.dictionary;
         while (dentry != NULL) {
             struct cos_dictionary_entry *odentry;
-            
+
             cos_free_object(dentry->key);
             cos_free_object(dentry->value);
-            
+
             odentry = dentry;
             dentry = dentry->next;
             free(odentry);
@@ -414,7 +415,48 @@ int cos_decode_number(struct pdf_doc *doc,
                       uint64_t *offset_out,
                       struct cos_object **cosobj_out)
 {
-    return -1;
+    struct cos_object *cosobj;
+    uint8_t c; /* current byte from source data */
+    int len; /* number of decimal places in number */
+    uint8_t num[21]; /* temporary buffer for decimal values */
+    uint64_t offset; /* current offset of source data */
+
+    offset = *offset_out;
+
+    for (len = 0; len < sizeof(num); len++) {
+        c = DOC_BYTE(doc, offset);
+        if ((bclass[c] & BC_DCML) != BC_DCML) {
+            int64_t result = 0; /* parsed result */
+            uint64_t tens;
+
+            if (len == 0) {
+                return -2; /* parse error no decimals in input */
+            }
+            /* sum value from each place */
+            for (tens = 1; len > 0; tens = tens * 10, len--) {
+                result += (num[len - 1] * tens);
+            }
+
+            doc_skip_ws(doc, &offset);
+
+            cosobj = calloc(1, sizeof(struct cos_object));
+            if (cosobj == NULL) {
+                return -1; /* memory error */
+            }
+
+            cosobj->type = COS_TYPE_INT;
+            cosobj->u.i = result;
+
+            *cosobj_out = cosobj;
+
+            *offset_out = offset;
+
+            return 0;
+        }
+        num[len] = c - '0';
+        offset++;
+    }
+    return -1; /* number too long */
 }
 
 int cos_decode_string(struct pdf_doc *doc,
@@ -459,19 +501,22 @@ int cos_decode_dictionary(struct pdf_doc *doc,
         return -1; /* memory error */
     }
     cosobj->type = COS_TYPE_DICTIONARY;
-    
+
     while ((DOC_BYTE(doc, offset) != '>') &&
            (DOC_BYTE(doc, offset + 1) != '>')) {
-        
+
         res = cos_decode_object(doc, &offset, &key);
         if (res != 0) {
             /* todo free up any dictionary entries already created */
+            printf("key object decode failed\n");
             return res;
         }
         if (key->type != COS_TYPE_NAME) {
             /* key value pairs without a name */
+            printf("key was %d not a name %d\n", key->type, COS_TYPE_NAME);
             return -1; /* syntax error */
         }
+        printf("key: %s\n", key->u.n);
 
         res = cos_decode_object(doc, &offset, &value);
         if (res != 0) {
@@ -491,7 +536,7 @@ int cos_decode_dictionary(struct pdf_doc *doc,
         entry->next = cosobj->u.dictionary;
 
         cosobj->u.dictionary = entry;
-        
+
     }
     offset += 2; /* skip closing >> */
     doc_skip_ws(doc, &offset);
@@ -511,6 +556,11 @@ int cos_decode_list(struct pdf_doc *doc,
 
 #define NAME_MAX_LENGTH 127
 
+/**
+ * decode a name object
+ *
+ * \todo deal with # symbols on pdf versions 1.2 and later
+ */
 int cos_decode_name(struct pdf_doc *doc,
                       uint64_t *offset_out,
                       struct cos_object **cosobj_out)
@@ -519,13 +569,31 @@ int cos_decode_name(struct pdf_doc *doc,
     struct cos_object *cosobj;
     uint8_t c;
     char name[NAME_MAX_LENGTH + 1];
+    int idx = 0;
 
     offset = *offset_out;
 
     c = DOC_BYTE(doc, offset++);
+    if (c != '/') {
+        return -1; /* names must be prefixed with a / */
+    }
+    printf("found a name\n");
 
+    c = DOC_BYTE(doc, offset++);
+    while ((idx <= NAME_MAX_LENGTH) &&
+           ((bclass[c] & (BC_WSPC | BC_DELM)) == 0)) {
+        //printf("%c", c);
+        name[idx++] = c;
+        c = DOC_BYTE(doc, offset++);
+    }
+    //printf("\nidx: %d\n", idx);
+    if (idx > NAME_MAX_LENGTH) {
+        /* name length exceeded implementation limit */
+        return -1;
+    }
+    name[idx] = 0;
 
-
+    //printf("name: %s\n", name);
 
     doc_skip_ws(doc, &offset);
 
@@ -534,8 +602,8 @@ int cos_decode_name(struct pdf_doc *doc,
         return -1; /* memory error */
     }
 
-    cosobj->type = COS_TYPE_BOOL;
-    cosobj->u.b = value;
+    cosobj->type = COS_TYPE_NAME;
+    cosobj->u.n = strdup(name);
 
     *cosobj_out = cosobj;
 
@@ -553,7 +621,7 @@ int cos_decode_boolean(struct pdf_doc *doc,
     struct cos_object *cosobj;
     uint8_t c;
     bool value;
-    
+
     offset = *offset_out;
 
     c = DOC_BYTE(doc, offset++);
@@ -573,7 +641,7 @@ int cos_decode_boolean(struct pdf_doc *doc,
             return -1; /* syntax error */
         }
         value = true;
-        
+
     } else if ((c == 'f') || (c == 'F')) {
         /* false branch */
 
@@ -613,7 +681,7 @@ int cos_decode_boolean(struct pdf_doc *doc,
     *cosobj_out = cosobj;
 
     *offset_out = offset;
-    
+
     return 0;
 
 }
@@ -625,7 +693,7 @@ int cos_decode_null(struct pdf_doc *doc,
     uint64_t offset;
     struct cos_object *cosobj;
     uint8_t c;
-    
+
     offset = *offset_out;
 
     c = DOC_BYTE(doc, offset++);
@@ -644,7 +712,7 @@ int cos_decode_null(struct pdf_doc *doc,
     if ((c != 'l') && (c != 'L')) {
         return -1; /* syntax error */
     }
-    
+
     doc_skip_ws(doc, &offset);
 
     cosobj = calloc(1, sizeof(struct cos_object));
@@ -658,6 +726,83 @@ int cos_decode_null(struct pdf_doc *doc,
     return 0;
 }
 
+/**
+ * attempt to decode the stream into a reference
+ *
+ * The stream has already had a positive integer decoded from it. if another
+ * positive integer follows and a R character after that it is a reference,
+ * otherwise bail, but not finding a ref is not an error!
+ *
+ * \param doc the pdf document
+ * \param offset_out offset of current cursor in stream
+ * \param cosobj_out the object to return into, on input contains the first
+ * integer
+ */
+int cos_attempt_decode_reference(struct pdf_doc *doc,
+                      uint64_t *offset_out,
+                      struct cos_object **cosobj_out)
+{
+    uint64_t offset;
+    struct cos_object *cosobj; /* possible generation object */
+    uint8_t c;
+    int res;
+    struct cos_reference *nref; /* new reference */
+
+    offset = *offset_out;
+
+    res = cos_decode_object(doc, &offset, &cosobj);
+    if (res != 0) {
+        return 0; /* no error if object could not be decoded */
+    }
+
+    if (cosobj->type != COS_TYPE_INT) {
+        /* next object was not an integer so not a reference */
+        cos_free_object(cosobj);
+        return 0;
+    }
+
+    if (cosobj->u.i < 0) {
+        /* integer was negative so not a reference (generations must be
+         * non-negative
+         */
+        cos_free_object(cosobj);
+        return 0;
+
+    }
+
+    /* two int in a row, look for the R */
+    c = DOC_BYTE(doc, offset++);
+    if (c != 'R') {
+        /* no R so not a reference */
+        cos_free_object(cosobj);
+        return 0;
+    }
+
+    /* found reference */
+
+    printf("found reference\n");
+    doc_skip_ws(doc, &offset);
+
+    nref = calloc(1, sizeof(struct cos_reference));
+    if (nref == NULL) {
+        /* todo free objects */
+        return -1; /* memory error */
+    }
+
+    nref->id = (*cosobj_out)->u.i;
+    nref->generation = cosobj->u.i;
+
+    cos_free_object(*cosobj_out);
+
+    cosobj->type = COS_TYPE_REFERENCE;
+    cosobj->u.reference = nref;
+
+    *cosobj_out = cosobj;
+
+    *offset_out = offset;
+
+    return 0;
+}
 
 /**
  * Decode input stream into an object
@@ -678,7 +823,7 @@ int cos_decode_null(struct pdf_doc *doc,
  * [ a list
  * t|T boolean true
  * f|F boolean false
- * n|N null 
+ * n|N null
  *
  * Grammar is:
  * cos_object:
@@ -711,12 +856,12 @@ int cos_decode_object(struct pdf_doc *doc,
     uint64_t offset;
     int res;
     struct cos_object *cosobj;
-    
+
     offset = *offset_out;
 
     /* object could be any type use first char to try and select */
     switch (DOC_BYTE(doc, offset)) {
-        
+
     case '-':
     case '+':
     case '.':
@@ -731,7 +876,12 @@ int cos_decode_object(struct pdf_doc *doc,
     case '8':
     case '9':
         res = cos_decode_number(doc, &offset, &cosobj);
-        /* if type is uint try to check for reference */
+        /* if type is positive integer try to check for reference */
+        if ((res == 0) &&
+            (cosobj->type == COS_TYPE_INT) &&
+            (cosobj->u.i > 0)) {
+            res = cos_attempt_decode_reference(doc, &offset, &cosobj);
+        }
         break;
 
     case '<':
@@ -769,7 +919,13 @@ int cos_decode_object(struct pdf_doc *doc,
     default:
         res = -1; /* syntax error */
     }
-    
+
+
+    if (res == 0) {
+        *cosobj_out = cosobj;
+        *offset_out = offset;
+    }
+
     return res;
 }
 
@@ -792,7 +948,7 @@ int decode_trailer(struct pdf_doc *doc, uint64_t offset)
     }
     offset += 7;
     doc_skip_ws(doc, &offset);
-    
+
     res = cos_decode_object(doc, &offset, &trailer);
     if (res != 0) {
         return res;
@@ -802,7 +958,7 @@ int decode_trailer(struct pdf_doc *doc, uint64_t offset)
         cos_free_object(trailer);
         return -1;
     }
-    
+
     return 0;
 }
 
@@ -812,9 +968,9 @@ int decode_xref(struct pdf_doc *doc, uint64_t *offset_out)
     uint64_t objnum; /* current object number */
     uint64_t lastobjnum;
     uint64_t offset;
-    
+
     offset = *offset_out;
-    
+
     /* xref object header */
     if ((DOC_BYTE(doc, offset    ) != 'x') &&
         (DOC_BYTE(doc, offset + 1) != 'r') &&
@@ -866,7 +1022,7 @@ int decode_xref(struct pdf_doc *doc, uint64_t *offset_out)
             objnum++;
         }
         //        printf("at objnum %"PRIu64"\n", objnum);
-        
+
         /* first object number in table */
         res = doc_read_uint(doc, &offset, &objnum);
     }
@@ -909,6 +1065,6 @@ int main(int argc, char **argv)
         printf("failed to decode trailer\n");
         return res;
     }
-    
+
     return 0;
 }