|
| 1 | +from lxml.html.clean import Cleaner |
| 2 | + |
| 3 | +from parsel import Selector |
| 4 | + |
| 5 | +HTML_BODY = """ |
| 6 | +<body> |
| 7 | + <div class="product"> |
| 8 | + <div class="name">Product1</div> |
| 9 | + <span class="price"><b>Price:</b>100</span> |
| 10 | + </div> |
| 11 | + <div class="product"> |
| 12 | + <div class="name">Product2</div> |
| 13 | + <span class="price"><b>Price:</b>200</span> |
| 14 | + </div> |
| 15 | +</body> |
| 16 | +""" |
| 17 | + |
| 18 | + |
| 19 | +def test_text_get() -> None: |
| 20 | + sel = Selector("<p>title:<h1>some text</h1></p>") |
| 21 | + txt = sel.get(text=True) |
| 22 | + assert txt == "title:\n\nsome text" |
| 23 | + |
| 24 | + |
| 25 | +def test_text_getall() -> None: |
| 26 | + sel = Selector("<ul><li>option1</li><li>option2</li></ul>") |
| 27 | + |
| 28 | + assert sel.getall(text=True) == ["option1\noption2"] |
| 29 | + assert sel.css("li").getall(text=True) == ["option1", "option2"] |
| 30 | + |
| 31 | + |
| 32 | +def test_cleaned() -> None: |
| 33 | + div_html = "<div><script>SCRIPT</script>" "<style>STYLE</style><p>hello</p><div>" |
| 34 | + sel = Selector(div_html) |
| 35 | + assert sel.css("script").getall() == ["<script>SCRIPT</script>"] |
| 36 | + assert sel.cleaned().css("script").getall() == [] |
| 37 | + |
| 38 | + assert len(sel.css("script")) == 1 |
| 39 | + assert len(sel.css("style")) == 1 |
| 40 | + assert len(sel.css("p")) == 1 |
| 41 | + |
| 42 | + assert len(sel.cleaned().css("script")) == 0 |
| 43 | + assert len(sel.cleaned().css("style")) == 1 |
| 44 | + assert len(sel.cleaned().css("p")) == 1 |
| 45 | + |
| 46 | + |
| 47 | +def test_cleaned_options() -> None: |
| 48 | + div_html = "<div><script>SCRIPT</script>" "<style>STYLE</style><p>hello</p><div>" |
| 49 | + sel = Selector(div_html) |
| 50 | + assert len(sel.css("script")) == 1 |
| 51 | + assert len(sel.css("style")) == 1 |
| 52 | + assert len(sel.css("p")) == 1 |
| 53 | + |
| 54 | + assert len(sel.cleaned().css("script")) == 0 |
| 55 | + assert len(sel.cleaned().css("style")) == 1 |
| 56 | + assert len(sel.cleaned().css("p")) == 1 |
| 57 | + |
| 58 | + assert len(sel.cleaned("html").css("script")) == 0 |
| 59 | + assert len(sel.cleaned("html").css("style")) == 1 |
| 60 | + assert len(sel.cleaned("html").css("p")) == 1 |
| 61 | + |
| 62 | + assert len(sel.cleaned("text").css("script")) == 0 |
| 63 | + assert len(sel.cleaned("text").css("style")) == 0 |
| 64 | + assert len(sel.cleaned("text").css("p")) == 1 |
| 65 | + |
| 66 | + cleaner = Cleaner(kill_tags=["p"], scripts=False, style=False) |
| 67 | + assert len(sel.cleaned(cleaner).css("script")) == 1 |
| 68 | + assert len(sel.cleaned(cleaner).css("style")) == 1 |
| 69 | + assert len(sel.cleaned(cleaner).css("p")) == 0 |
| 70 | + |
| 71 | + |
| 72 | +def test_get_cleaner() -> None: |
| 73 | + div_html = "<div><script>SCRIPT</script><style>STYLE</style><p>P</p></div>" |
| 74 | + sel = Selector(div_html) |
| 75 | + cleaner = Cleaner(kill_tags=["p"], scripts=False, style=False) |
| 76 | + |
| 77 | + assert sel.get(text=True) == "P" |
| 78 | + assert sel.get(text=True, cleaner=None) == "SCRIPT STYLE\n\nP" |
| 79 | + assert sel.get(text=True, cleaner="html") == "STYLE\n\nP" |
| 80 | + assert sel.get(text=True, cleaner="text") == "P" |
| 81 | + assert sel.get(text=True, cleaner=cleaner) == "SCRIPT STYLE" |
| 82 | + |
| 83 | + div = sel.css("div") |
| 84 | + assert div.get() == div_html |
| 85 | + assert div.get(cleaner=None) == div_html |
| 86 | + assert div.get(cleaner="html") == "<div><style>STYLE</style><p>P</p></div>" |
| 87 | + assert div.get(cleaner="text") == "<div><p>P</p></div>" |
| 88 | + assert ( |
| 89 | + div.get(cleaner=cleaner) |
| 90 | + == "<div><script>SCRIPT</script><style>STYLE</style></div>" |
| 91 | + ) |
| 92 | + |
| 93 | + |
| 94 | +def test_guess_punct_space() -> None: |
| 95 | + sel = Selector('<p>hello<b>"Folks"</b></p>') |
| 96 | + assert sel.get(text=True, guess_punct_space=False) == 'hello "Folks"' |
| 97 | + assert sel.get(text=True, guess_punct_space=True) == 'hello"Folks"' |
| 98 | + |
| 99 | + assert sel.getall(text=True, guess_punct_space=False) == ['hello "Folks"'] |
| 100 | + assert sel.getall(text=True, guess_punct_space=True) == ['hello"Folks"'] |
| 101 | + |
| 102 | + |
| 103 | +def test_guess_layout() -> None: |
| 104 | + sel = Selector("<ul><li>option1</li><li>option2</li></ul>") |
| 105 | + assert sel.get(text=True, guess_layout=False) == "option1 option2" |
| 106 | + assert sel.get(text=True, guess_layout=True) == "option1\noption2" |
| 107 | + |
| 108 | + assert sel.getall(text=True, guess_layout=False) == ["option1 option2"] |
| 109 | + assert sel.getall(text=True, guess_layout=True) == ["option1\noption2"] |
0 commit comments