1<h2>mxj - to/from maps, XML and JSON</h2>
2Decode/encode XML to/from map[string]interface{} (or JSON) values, and extract/modify values from maps by key or key-path, including wildcards.
3
4mxj supplants the legacy x2j and j2x packages. If you want the old syntax, use mxj/x2j and mxj/j2x packages.
5
6<h4>Installation</h4>
7Using go.mod:
8<pre>
9go get github.com/clbanning/mxj/v2@v2.7
10</pre>
11
12<pre>
13import "github.com/clbanning/mxj/v2"
14</pre>
15
16... or just vendor the package.
17
18<h4>Related Packages</h4>
19
20https://github.com/clbanning/checkxml provides functions for validating XML data.
21
22<h4>Refactor Encoder - 2020.05.01</h4>
23Issue #70 highlighted that encoding large maps does not scale well, since the original logic used string appends operations. Using bytes.Buffer results in linear scaling for very large XML docs. (Metrics based on MacBook Pro i7 w/ 16 GB.)
24
25 Nodes m.XML() time
26 54809 12.53708ms
27 109780 32.403183ms
28 164678 59.826412ms
29 482598 109.358007ms
30
31<h4>Refactor Decoder - 2015.11.15</h4>
32For over a year I've wanted to refactor the XML-to-map[string]interface{} decoder to make it more performant. I recently took the time to do that, since we were using github.com/clbanning/mxj in a production system that could be deployed on a Raspberry Pi. Now the decoder is comparable to the stdlib JSON-to-map[string]interface{} decoder in terms of its additional processing overhead relative to decoding to a structure value. As shown by:
33
34 BenchmarkNewMapXml-4 100000 18043 ns/op
35 BenchmarkNewStructXml-4 100000 14892 ns/op
36 BenchmarkNewMapJson-4 300000 4633 ns/op
37 BenchmarkNewStructJson-4 300000 3427 ns/op
38 BenchmarkNewMapXmlBooks-4 20000 82850 ns/op
39 BenchmarkNewStructXmlBooks-4 20000 67822 ns/op
40 BenchmarkNewMapJsonBooks-4 100000 17222 ns/op
41 BenchmarkNewStructJsonBooks-4 100000 15309 ns/op
42
43<h4>Notices</h4>
44
45 2022.11.28: v2.7 - add SetGlobalKeyMapPrefix to change default prefix, '#', for default keys
46 2022.11.20: v2.6 - add NewMapForattedXmlSeq for XML docs formatted with whitespace character
47 2021.02.02: v2.5 - add XmlCheckIsValid toggle to force checking that the encoded XML is valid
48 2020.12.14: v2.4 - add XMLEscapeCharsDecoder to preserve XML escaped characters in Map values
49 2020.10.28: v2.3 - add TrimWhiteSpace option
50 2020.05.01: v2.2 - optimize map to XML encoding for large XML docs.
51 2019.07.04: v2.0 - remove unnecessary methods - mv.XmlWriterRaw, mv.XmlIndentWriterRaw - for Map and MapSeq.
52 2019.07.04: Add MapSeq type and move associated functions and methods from Map to MapSeq.
53 2019.01.21: DecodeSimpleValuesAsMap - decode to map[<tag>:map["#text":<value>]] rather than map[<tag>:<value>]
54 2018.04.18: mv.Xml/mv.XmlIndent encodes non-map[string]interface{} map values - map[string]string, map[int]uint, etc.
55 2018.03.29: mv.Gob/NewMapGob support gob encoding/decoding of Maps.
56 2018.03.26: Added mxj/x2j-wrapper sub-package for migrating from legacy x2j package.
57 2017.02.22: LeafNode paths can use ".N" syntax rather than "[N]" for list member indexing.
58 2017.02.10: SetFieldSeparator changes field separator for args in UpdateValuesForPath, ValuesFor... methods.
59 2017.02.06: Support XMPP stream processing - HandleXMPPStreamTag().
60 2016.11.07: Preserve name space prefix syntax in XmlSeq parser - NewMapXmlSeq(), etc.
61 2016.06.25: Support overriding default XML attribute prefix, "-", in Map keys - SetAttrPrefix().
62 2016.05.26: Support customization of xml.Decoder by exposing CustomDecoder variable.
63 2016.03.19: Escape invalid chars when encoding XML attribute and element values - XMLEscapeChars().
64 2016.03.02: By default decoding XML with float64 and bool value casting will not cast "NaN", "Inf", and "-Inf".
65 To cast them to float64, first set flag with CastNanInf(true).
66 2016.02.22: New mv.Root(), mv.Elements(), mv.Attributes methods let you examine XML document structure.
67 2016.02.16: Add CoerceKeysToLower() option to handle tags with mixed capitalization.
68 2016.02.12: Seek for first xml.StartElement token; only return error if io.EOF is reached first (handles BOM).
69 2015.12.02: XML decoding/encoding that preserves original structure of document. See NewMapXmlSeq()
70 and mv.XmlSeq() / mv.XmlSeqIndent().
71 2015-05-20: New: mv.StringIndentNoTypeInfo().
72 Also, alphabetically sort map[string]interface{} values by key to prettify output for mv.Xml(),
73 mv.XmlIndent(), mv.StringIndent(), mv.StringIndentNoTypeInfo().
74 2014-11-09: IncludeTagSeqNum() adds "_seq" key with XML doc positional information.
75 (NOTE: PreserveXmlList() is similar and will be here soon.)
76 2014-09-18: inspired by NYTimes fork, added PrependAttrWithHyphen() to allow stripping hyphen from attribute tag.
77 2014-08-02: AnyXml() and AnyXmlIndent() will try to marshal arbitrary values to XML.
78 2014-04-28: ValuesForPath() and NewMap() now accept path with indexed array references.
79
80<h4>Basic Unmarshal XML to map[string]interface{}</h4>
81<pre>type Map map[string]interface{}</pre>
82
83Create a `Map` value, 'mv', from any `map[string]interface{}` value, 'v':
84<pre>mv := Map(v)</pre>
85
86Unmarshal / marshal XML as a `Map` value, 'mv':
87<pre>mv, err := NewMapXml(xmlValue) // unmarshal
88xmlValue, err := mv.Xml() // marshal</pre>
89
90Unmarshal XML from an `io.Reader` as a `Map` value, 'mv':
91<pre>mv, err := NewMapXmlReader(xmlReader) // repeated calls, as with an os.File Reader, will process stream
92mv, raw, err := NewMapXmlReaderRaw(xmlReader) // 'raw' is the raw XML that was decoded</pre>
93
94Marshal `Map` value, 'mv', to an XML Writer (`io.Writer`):
95<pre>err := mv.XmlWriter(xmlWriter)
96raw, err := mv.XmlWriterRaw(xmlWriter) // 'raw' is the raw XML that was written on xmlWriter</pre>
97
98Also, for prettified output:
99<pre>xmlValue, err := mv.XmlIndent(prefix, indent, ...)
100err := mv.XmlIndentWriter(xmlWriter, prefix, indent, ...)
101raw, err := mv.XmlIndentWriterRaw(xmlWriter, prefix, indent, ...)</pre>
102
103Bulk process XML with error handling (note: handlers must return a boolean value):
104<pre>err := HandleXmlReader(xmlReader, mapHandler(Map), errHandler(error))
105err := HandleXmlReaderRaw(xmlReader, mapHandler(Map, []byte), errHandler(error, []byte))</pre>
106
107Converting XML to JSON: see Examples for `NewMapXml` and `HandleXmlReader`.
108
109There are comparable functions and methods for JSON processing.
110
111Arbitrary structure values can be decoded to / encoded from `Map` values:
112<pre>mv, err := NewMapStruct(structVal)
113err := mv.Struct(structPointer)</pre>
114
115<h4>Extract / modify Map values</h4>
116To work with XML tag values, JSON or Map key values or structure field values, decode the XML, JSON
117or structure to a `Map` value, 'mv', or cast a `map[string]interface{}` value to a `Map` value, 'mv', then:
118<pre>paths := mv.PathsForKey(key)
119path := mv.PathForKeyShortest(key)
120values, err := mv.ValuesForKey(key, subkeys)
121values, err := mv.ValuesForPath(path, subkeys)
122count, err := mv.UpdateValuesForPath(newVal, path, subkeys)</pre>
123
124Get everything at once, irrespective of path depth:
125<pre>leafnodes := mv.LeafNodes()
126leafvalues := mv.LeafValues()</pre>
127
128A new `Map` with whatever keys are desired can be created from the current `Map` and then encoded in XML
129or JSON. (Note: keys can use dot-notation.)
130<pre>newMap, err := mv.NewMap("oldKey_1:newKey_1", "oldKey_2:newKey_2", ..., "oldKey_N:newKey_N")
131newMap, err := mv.NewMap("oldKey1", "oldKey3", "oldKey5") // a subset of 'mv'; see "examples/partial.go"
132newXml, err := newMap.Xml() // for example
133newJson, err := newMap.Json() // ditto</pre>
134
135<h4>Usage</h4>
136
137The package is fairly well [self-documented with examples](http://godoc.org/github.com/clbanning/mxj).
138
139Also, the subdirectory "examples" contains a wide range of examples, several taken from golang-nuts discussions.
140
141<h4>XML parsing conventions</h4>
142
143Using NewMapXml()
144
145 - Attributes are parsed to `map[string]interface{}` values by prefixing a hyphen, `-`,
146 to the attribute label. (Unless overridden by `PrependAttrWithHyphen(false)` or
147 `SetAttrPrefix()`.)
148 - If the element is a simple element and has attributes, the element value
149 is given the key `#text` for its `map[string]interface{}` representation. (See
150 the 'atomFeedString.xml' test data, below.)
151 - XML comments, directives, and process instructions are ignored.
152 - If CoerceKeysToLower() has been called, then the resultant keys will be lower case.
153
154Using NewMapXmlSeq()
155
156 - Attributes are parsed to `map["#attr"]map[<attr_label>]map[string]interface{}`values
157 where the `<attr_label>` value has "#text" and "#seq" keys - the "#text" key holds the
158 value for `<attr_label>`.
159 - All elements, except for the root, have a "#seq" key.
160 - Comments, directives, and process instructions are unmarshalled into the Map using the
161 keys "#comment", "#directive", and "#procinst", respectively. (See documentation for more
162 specifics.)
163 - Name space syntax is preserved:
164 - `<ns:key>something</ns.key>` parses to `map["ns:key"]interface{}{"something"}`
165 - `xmlns:ns="http://myns.com/ns"` parses to `map["xmlns:ns"]interface{}{"http://myns.com/ns"}`
166
167Both
168
169 - By default, "Nan", "Inf", and "-Inf" values are not cast to float64. If you want them
170 to be cast, set a flag to cast them using CastNanInf(true).
171
172<h4>XML encoding conventions</h4>
173
174 - 'nil' `Map` values, which may represent 'null' JSON values, are encoded as `<tag/>`.
175 NOTE: the operation is not symmetric as `<tag/>` elements are decoded as `tag:""` `Map` values,
176 which, then, encode in JSON as `"tag":""` values.
177 - ALSO: there is no guarantee that the encoded XML doc will be the same as the decoded one. (Go
178 randomizes the walk through map[string]interface{} values.) If you plan to re-encode the
179 Map value to XML and want the same sequencing of elements look at NewMapXmlSeq() and
180 mv.XmlSeq() - these try to preserve the element sequencing but with added complexity when
181 working with the Map representation.
182
183<h4>Running "go test"</h4>
184
185Because there are no guarantees on the sequence map elements are retrieved, the tests have been
186written for visual verification in most cases. One advantage is that you can easily use the
187output from running "go test" as examples of calling the various functions and methods.
188
189<h4>Motivation</h4>
190
191I make extensive use of JSON for messaging and typically unmarshal the messages into
192`map[string]interface{}` values. This is easily done using `json.Unmarshal` from the
193standard Go libraries. Unfortunately, many legacy solutions use structured
194XML messages; in those environments the applications would have to be refactored to
195interoperate with my components.
196
197The better solution is to just provide an alternative HTTP handler that receives
198XML messages and parses it into a `map[string]interface{}` value and then reuse
199all the JSON-based code. The Go `xml.Unmarshal()` function does not provide the same
200option of unmarshaling XML messages into `map[string]interface{}` values. So I wrote
201a couple of small functions to fill this gap and released them as the x2j package.
202
203Over the next year and a half additional features were added, and the companion j2x
204package was released to address XML encoding of arbitrary JSON and `map[string]interface{}`
205values. As part of a refactoring of our production system and looking at how we had been
206using the x2j and j2x packages we found that we rarely performed direct XML-to-JSON or
207JSON-to_XML conversion and that working with the XML or JSON as `map[string]interface{}`
208values was the primary value. Thus, everything was refactored into the mxj package.
209
View as plain text