Project

General

Profile

1 13 aaronmk
# XML-database conversion
2
3 2418 aaronmk
import copy
4 16 aaronmk
import re
5 13 aaronmk
from xml.dom import Node
6
7 446 aaronmk
import exc
8 2436 aaronmk
import Parser
9 46 aaronmk
import sql
10 3077 aaronmk
import sql_io
11 2268 aaronmk
import sql_gen
12 84 aaronmk
import strings
13 133 aaronmk
import util
14 46 aaronmk
import xml_dom
15 2113 aaronmk
import xml_func
16 2436 aaronmk
import xpath
17 13 aaronmk
18 139 aaronmk
def name_of(node): return re.sub(r'^.*\.', r'', node.tagName)
19 16 aaronmk
20 13 aaronmk
ptr_suffix = '_id'
21
22 138 aaronmk
def is_ptr(node_name): return node_name.lower().endswith(ptr_suffix)
23 13 aaronmk
24 172 aaronmk
def ptr_type_guess(node_name):
25 13 aaronmk
    assert is_ptr(node_name)
26
    return node_name[:-len(ptr_suffix)]
27
28
def ptr_target(node):
29
    assert is_ptr(name_of(node))
30 2113 aaronmk
    return xml_dom.value_node(node)
31 13 aaronmk
32
def find_by_name(node, name):
33 46 aaronmk
    for parent in xml_dom.NodeParentIter(node):
34 13 aaronmk
        if name_of(parent) == name: return parent
35
        else:
36 46 aaronmk
            for child in xml_dom.NodeElemIter(parent):
37 16 aaronmk
                child_name = name_of(child)
38 13 aaronmk
                if is_ptr(child_name):
39 172 aaronmk
                    target = ptr_target(child)
40
                    if target.tagName == name: return target
41 13 aaronmk
                elif child_name == name: return child
42
    return None
43
44 2039 aaronmk
class ColRef:
45
    '''A reference to a table column'''
46
    def __init__(self, name, idx):
47
        self.name = name
48
        self.idx = idx
49
50
    def __str__(self): return self.name
51
52 3652 aaronmk
put_special_funcs = set(['_simplifyPath'])
53 2549 aaronmk
54 3650 aaronmk
def put(db, node, row_ins_ct_ref=None, on_error=exc.raise_,
55 3649 aaronmk
    col_defaults={}, in_table=None, parent_ids_loc=None, next=None):
56 3650 aaronmk
    def put_(node, in_row_ct_ref=None):
57
        return put(db, node, row_ins_ct_ref, on_error, col_defaults,
58 3649 aaronmk
            in_table, parent_ids_loc, next)
59 3621 aaronmk
60 3646 aaronmk
    def on_error_(e):
61
        exc.add_msg(e, 'node:\n'+str(node))
62
        on_error(e)
63
64 2551 aaronmk
    is_func = xml_func.is_func(node)
65 2005 aaronmk
    out_table = name_of(node)
66 2432 aaronmk
67
    # Divide children into fields and children with fkeys to parent
68 2005 aaronmk
    row = {}
69
    children = []
70
    for child in xml_dom.NodeElemIter(node):
71
        child_name = name_of(child)
72
        if xml_dom.is_empty(child): row[child_name] = None
73
        elif xml_dom.is_text(child):
74
            row[child_name] = strings.to_unicode(xml_dom.value(child))
75 2113 aaronmk
        else:
76
            child_value = xml_dom.value_node(child)
77 2434 aaronmk
            if is_func or is_ptr(child_name) or xml_func.is_func(child_value):
78 2432 aaronmk
                row[child_name] = child_value
79 2113 aaronmk
            else: children.append(child)
80 2434 aaronmk
81
    # Special handling for structural XML functions
82
    if out_table == '_simplifyPath':
83 2436 aaronmk
        # Parse args
84
        def wrap_e(e): raise xml_func.SyntaxError(e)
85 2434 aaronmk
        try:
86 3650 aaronmk
            next = row['next'] # modifies outer next var used by put_()
87 2434 aaronmk
            require = row['require']
88
            path = row['path']
89 2436 aaronmk
        except KeyError, e: wrap_e(e)
90
        try: next = xpath.parse(next)
91
        except Parser.SyntaxError, e: wrap_e(e)
92
        try: next = next[0].name
93
        except IndexError, e: wrap_e(e)
94 2434 aaronmk
95 3650 aaronmk
        return put_(path)
96 2434 aaronmk
97 3624 aaronmk
    in_tables = []
98
    no_empty = set()
99
    if in_table != None:
100
        in_tables.append(in_table)
101
        no_empty.add(in_table)
102 2966 aaronmk
103 3107 aaronmk
    def pkey(table): return sql.pkey(db, table, True)
104
105 2177 aaronmk
    # Add fkey to parent
106
    if parent_ids_loc != None:
107 3601 aaronmk
        if sql_gen.is_table_col(parent_ids_loc):
108
            no_empty.add(parent_ids_loc.table)
109 2177 aaronmk
        parent_ptr = node.getAttribute('fkey')
110
        if parent_ptr == '': parent_ptr = pkey(name_of(node.parentNode))
111
        row[parent_ptr] = parent_ids_loc
112
113 2060 aaronmk
    # Divide fields into input columns and literal values
114 2506 aaronmk
    parent_ids_loc = None # applies to this section
115 2060 aaronmk
    for out_col, value in row.iteritems():
116 2432 aaronmk
        # Handle forward pointers
117 3650 aaronmk
        if xml_dom.is_node(value): row[out_col] = value = put_(value)
118 2432 aaronmk
119
        # Translate values
120 2268 aaronmk
        if isinstance(value, sql_gen.Col): # value is temp table column
121
            in_tables.append(value.table)
122 2177 aaronmk
        elif util.is_str(value) and value.startswith(input_col_prefix):
123
            # value is input column
124 2272 aaronmk
            row[out_col] = sql_gen.Col(strings.remove_prefix(input_col_prefix,
125 2313 aaronmk
                value), in_table)
126 3428 aaronmk
        else: # value is literal value
127 2323 aaronmk
            row[out_col] = sql_gen.NamedCol(out_col, value)
128 2060 aaronmk
129
    # Insert node
130 3077 aaronmk
    pkeys_loc = sql_io.put_table(db, out_table, in_tables, row, row_ins_ct_ref,
131 3646 aaronmk
        next, col_defaults, on_error_)
132 2060 aaronmk
133 2975 aaronmk
    sql.empty_temp(db, set(in_tables) - no_empty)
134 2966 aaronmk
135 2177 aaronmk
    # Insert children with fkeys to parent
136 2506 aaronmk
    parent_ids_loc = pkeys_loc # applies to this section
137 3650 aaronmk
    for child in children: put_(child)
138 2177 aaronmk
139 2133 aaronmk
    return pkeys_loc
140 3651 aaronmk
141
def get(db, node, limit=None, start=None):
142
    def pkey(table): return sql.pkey(db, table)
143
144
    node = node.firstChild
145
    table = name_of(node)
146
    pkey_ = pkey(table)
147
148
    fields = []
149
    conds = {}
150
    for child in xml_dom.NodeElemIter(node):
151
        child_name = name_of(child)
152
        if xml_dom.is_empty(child): fields.append(child_name)
153
        elif xml_dom.is_text(child): conds[child_name] = xml_dom.value(child)
154
        else: raise Exception('Joins not supported yet')
155
    id_ = xml_dom.get_id(node)
156
    if id_ != None: conds[pkey(table)] = id_ # replace any existing pkey value
157
    if fields == []: fields.append(pkey_)
158
159
    return sql.select(db, table, fields, conds, limit, start)
160
161
# Controls when and how put_table() will partition the input table
162
partition_size = 500000 # rows
163
164
input_col_prefix = '$'
165
166
def put_table(db, node, in_table, in_row_ct_ref=None, row_ins_ct_ref=None,
167
    limit=None, start=0, on_error=exc.raise_, col_defaults={}):
168
    '''
169
    @param node The XML tree that transforms the input to the output. Similar to
170
        put()'s node param, but with the input column name prefixed by
171
        input_col_prefix in place of the column value.
172
    @return sql_gen.Col Where the pkeys (from INSERT RETURNING) are made
173
        available
174
    '''
175
    in_table = sql_gen.as_Table(in_table)
176
    in_table.set_srcs([in_table], overwrite=False)
177
    db.src = str(in_table)
178
179
    db.autoanalyze = True # but don't do this in row-based import
180
    db.autoexplain = True # but don't do this in row-based import
181
182
    # Import col_defaults
183
    for col, node_ in col_defaults.items():
184
        col_defaults[col] = put(db, node_, row_ins_ct_ref, on_error)
185
186
    # Subset and partition in_table
187
    # OK to do even if table already the right size because it takes <1 sec.
188
    full_in_table = in_table
189
    total = 0
190
    while limit == None or total < limit:
191
        # Adjust partition size if last partition
192
        this_limit = partition_size
193
        if limit != None: this_limit = min(this_limit, limit - total)
194
195
        # Row # is interally 0-based, but 1-based to the user
196
        db.log_debug('********** Partition: rows '+str(start+1)+'-'
197
            +str(start+this_limit)+' **********', level=1.2)
198
199
        # Subset in_table
200
        in_table = copy.copy(full_in_table) # don't modify input!
201
        in_table.name = str(in_table) # prepend schema
202
        cur = sql.run_query_into(db, sql.mk_select(db, full_in_table,
203
            limit=this_limit, start=start), into=in_table, add_pkey_=True)
204
            # full_in_table will be shadowed (hidden) by created temp table
205
206
        this_ct = cur.rowcount
207
        total += this_ct
208
        start += this_ct # advance start to fetch next set
209
        if this_ct == 0: break # in_table size is multiple of partition_size
210
211
        # Recurse
212
        pkeys_loc = put(db, node, row_ins_ct_ref, on_error,
213
            col_defaults, in_table)
214
        if in_row_ct_ref != None: in_row_ct_ref[0] += this_ct
215
216
        sql.empty_temp(db, in_table)
217
218
        if this_ct < partition_size: break # partial partition = last
219
220
        # Work around PostgreSQL's temp table disk space leak
221
        db.reconnect()
222
223
    return pkeys_loc